scripts/gdb: initial clk support: lx-clk-summary

Add an lx-clk-summary command which prints a subset of
/sys/kernel/debug/clk/clk_summary.

This can be used to examine hangs caused by clk not being enabled.

Link: http://lkml.kernel.org/r/Message-ID:
Signed-off-by: Leonard Crestez <leonard.crestez@nxp.com>
Cc: Jan Kiszka <jan.kiszka@siemens.com>
Cc: Jason Wessel <jason.wessel@windriver.com>
Cc: Kieran Bingham <kbingham@kernel.org>
Cc: Stephen Boyd <swboyd@chromium.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

scripts/gdb: add hlist utilities

This allows easily examining kernel hlists in python.

Link: http://lkml.kernel.org/r/Message-ID:
Signed-off-by: Leonard Crestez <leonard.crestez@nxp.com>
Reviewed-by: Stephen Boyd <sboyd@kernel.org>
Cc: Jason Wessel <jason.wessel@windriver.com>
Cc: Jan Kiszka <jan.kiszka@siemens.com>
Cc: Kieran Bingham <kbingham@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

scripts/gdb: silence pep8 checks

These scripts have some pep8 style warnings.  Fix them up so that this
directory is all pep8 clean.

Link: http://lkml.kernel.org/r/20190329220844.38234-6-swboyd@chromium.org
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Cc: Douglas Anderson <dianders@chromium.org>
Cc: Nikolay Borisov <n.borisov.lkml@gmail.com>
Cc: Kieran Bingham <kbingham@kernel.org>
Cc: Jan Kiszka <jan.kiszka@siemens.com>
Cc: Jackie Liu <liuyun01@kylinos.cn>
Cc: Jason Wessel <jason.wessel@windriver.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

scripts/gdb: add a timer list command

Implement a command to print the timer list, much like how
/proc/timer_list is implemented.  This can be used to look at the
pending timers on a crashed system.

[swboyd@chromium.org: v2]
Link: http://lkml.kernel.org/r/20190329220844.38234-5-swboyd@chromium.org
Link: http://lkml.kernel.org/r/20190325184522.260535-5-swboyd@chromium.org
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Cc: Douglas Anderson <dianders@chromium.org>
Cc: Nikolay Borisov <n.borisov.lkml@gmail.com>
Cc: Kieran Bingham <kbingham@kernel.org>
Cc: Jan Kiszka <jan.kiszka@siemens.com>
Cc: Jackie Liu <liuyun01@kylinos.cn>
Cc: Jason Wessel <jason.wessel@windriver.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

scripts/gdb: add rb tree iterating utilities

Implement gdb functions for rb_first(), rb_last(), rb_next(), and
rb_prev().  These can be useful to iterate through the kernel's
red-black trees.

[swboyd@chromium.org: v2]
Link: http://lkml.kernel.org/r/20190329220844.38234-4-swboyd@chromium.org
Link: http://lkml.kernel.org/r/20190325184522.260535-4-swboyd@chromium.org
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Cc: Douglas Anderson <dianders@chromium.org>
Cc: Nikolay Borisov <n.borisov.lkml@gmail.com>
Cc: Kieran Bingham <kbingham@kernel.org>
Cc: Jan Kiszka <jan.kiszka@siemens.com>
Cc: Jackie Liu <liuyun01@kylinos.cn>
Cc: Jason Wessel <jason.wessel@windriver.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

scripts/gdb: add kernel config dumping command

lx-configdump <file> dumps the contents of the gzipped .config to a text
file when the config is included in the kernel with CONFIG_IKCONFIG.  By
default, the file written is called config.txt, but it can be any user
supplied filename as well.  If the kernel config is in a module
(configs.ko), then it can be loaded along with symbols for the module
loaded with 'lx-symbols' and then this command will still work.

Obviously if you have the whole vmlinux then this can also be achieved
with scripts/extract-ikconfig, but this gdb script can be useful to
confirm that the memory contents of the config in memory and the vmlinux
contents on disk match what is expected.

[swboyd@chromium.org: v2]
Link: http://lkml.kernel.org/r/20190329220844.38234-3-swboyd@chromium.org
Link: http://lkml.kernel.org/r/20190325184522.260535-3-swboyd@chromium.org
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Cc: Douglas Anderson <dianders@chromium.org>
Cc: Nikolay Borisov <n.borisov.lkml@gmail.com>
Cc: Kieran Bingham <kbingham@kernel.org>
Cc: Jan Kiszka <jan.kiszka@siemens.com>
Cc: Jackie Liu <liuyun01@kylinos.cn>
Cc: Jason Wessel <jason.wessel@windriver.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

scripts/gdb: find vmlinux where it was before

Patch series "gdb script for kconfig and timer list".

This is a handful of changes to the kernel's gdb scripts to do some more
debugging with kgdb.  The first patch allows the vmlinux to be reloaded
from where it was specified on the command line so that this set of
scripts can be used from anywhere.  The second patch adds a script to
dump the config.gz to a file on the host debugging machine.  The third
patch adds some rb tree utilities and the last patch uses those rb tree
walking utilities to dump out the contents of /proc/timer_list from a
system under debug.

This patch (of 5):

If I run 'gdb <path/to/vmlinux>' and there's the vmlinux-gdb.py file
there I can properly see symbols and use the lx commands provided by the
GDB scripts.  But once I run 'lx-symbols' at the command prompt, gdb
reloads the vmlinux symbols assuming that this script was run from the
directory that has vmlinux at the root.  That isn't always true, but we
could just look and see what symbols were already loaded and use that
instead.  Let's do that so this can work by being invoked anywhere.

Link: http://lkml.kernel.org/r/20190325184522.260535-2-swboyd@chromium.org
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Cc: Douglas Anderson <dianders@chromium.org>
Cc: Nikolay Borisov <n.borisov.lkml@gmail.com>
Cc: Kieran Bingham <kbingham@kernel.org>
Cc: Jan Kiszka <jan.kiszka@siemens.com>
Cc: Jackie Liu <liuyun01@kylinos.cn>
Cc: Jason Wessel <jason.wessel@windriver.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

pps: pps-gpio PPS ECHO implementation

This patch implements the PPS ECHO functionality for pps-gpio, that
sysfs claims is available already.

Configuration is done via device tree bindings.

No changes are made to userspace interfaces.

This patch was originally written by Lukas Senger as part of a masters
thesis project and modified for inclusion into the linux kernel by Tom
Burkart.

Link: http://lkml.kernel.org/r/20190324043305.6627-4-tom@aussec.com
Signed-off-by: Tom Burkart <tom@aussec.com>
Acked-by: Rodolfo Giometti <giometti@enneenne.com>
Signed-off-by: Lukas Senger <lukas@fridolin.com>
Cc: Philipp Zabel <philipp.zabel@gmail.com>
Cc: Rob Herring <robh@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

dt-bindings: pps: pps-gpio PPS ECHO implementation

This patch implements the device tree binding changes required for the
PPS ECHO functionality for pps-gpio, that sysfs claims is available
already.

It adds two DT properties for configuring the PPS ECHO functionality.

This patch is provided separated from the rest of the patch per
Documentation/devicetree/bindings/submitting-patches.txt.

This patch was originally written by Lukas Senger as part of a masters
thesis project and modified for inclusion into the linux kernel by Tom
Burkart.

Link: http://lkml.kernel.org/r/20190324043305.6627-3-tom@aussec.com
Signed-off-by: Tom Burkart <tom@aussec.com>
Signed-off-by: Lukas Senger <lukas@fridolin.com>
Acked-by: Rodolfo Giometti <giometti@enneenne.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Cc: Philipp Zabel <philipp.zabel@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

pps: descriptor-based gpio

This patch changes the GPIO access for the pps-gpio driver from the
integer based API to the descriptor based API.

The integer based API is considered deprecated and the descriptor based
API is the preferred way to access GPIOs as per
Documentation/driver-api/gpio/intro.rst

No changes are made to userspace interfaces.

Link: http://lkml.kernel.org/r/20190324043305.6627-2-tom@aussec.com
Signed-off-by: Tom Burkart <tom@aussec.com>
Acked-by: Rodolfo Giometti <giometti@enneenne.com>
Reviewed-by: Philipp Zabel <philipp.zabel@gmail.com>
Cc: Lukas Senger <lukas@fridolin.com>
Cc: Rob Herring <robh@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

panic/reboot: allow specifying reboot_mode for panic only

Allow specifying reboot_mode for panic only.  This is needed on systems
where ramoops is used to store panic logs, and user wants to use warm
reset to preserve those, while still having cold reset on normal
reboots.

Link: http://lkml.kernel.org/r/20190322004735.27702-1-aaro.koskinen@iki.fi
Signed-off-by: Aaro Koskinen <aaro.koskinen@nokia.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

panic: avoid the extra noise dmesg

When kernel panic happens, it will first print the panic call stack,
then the ending msg like:

[   35.743249] ---[ end Kernel panic - not syncing: Fatal exception
[   35.749975] ------------[ cut here ]------------

The above message are very useful for debugging.

But if system is configured to not reboot on panic, say the
"panic_timeout" parameter equals 0, it will likely print out many noisy
message like WARN() call stack for each and every CPU except the panic
one, messages like below:

WARNING: CPU: 1 PID: 280 at kernel/sched/core.c:1198 set_task_cpu+0x183/0x190
Call Trace:
<IRQ>
try_to_wake_up
default_wake_function
autoremove_wake_function
__wake_up_common
__wake_up_common_lock
__wake_up
wake_up_klogd_work_func
irq_work_run_list
irq_work_tick
update_process_times
tick_sched_timer
__hrtimer_run_queues
hrtimer_interrupt
smp_apic_timer_interrupt
apic_timer_interrupt

For people working in console mode, the screen will first show the panic
call stack, but immediately overridden by these noisy extra messages,
which makes debugging much more difficult, as the original context gets
lost on screen.

Also these noisy messages will confuse some users, as I have seen many bug
reporters posted the noisy message into bugzilla, instead of the real
panic call stack and context.

Adding a flag "suppress_printk" which gets set in panic() to avoid those
noisy messages, without changing current kernel behavior that both panic
blinking and sysrq magic key can work as is, suggested by Petr Mladek.

To verify this, make sure kernel is not configured to reboot on panic and
in console
 # echo c > /proc/sysrq-trigger
to see if console only prints out the panic call stack.

Link: http://lkml.kernel.org/r/1551430186-24169-1-git-send-email-feng.tang@intel.com
Signed-off-by: Feng Tang <feng.tang@intel.com>
Suggested-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: Petr Mladek <pmladek@suse.com>
Acked-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Acked-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Kees Cook <keescook@chromium.org>
Cc: Borislav Petkov <bp@suse.de>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Jiri Slaby <jslaby@suse.com>
Cc: Sasha Levin <sashal@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

gcov: clang support

LLVM uses profiling data that's deliberately similar to GCC, but has a
very different way of exporting that data.  LLVM calls llvm_gcov_init()
once per module, and provides a couple of callbacks that we can use to
ask for more data.

We care about the "writeout" callback, which in turn calls back into
compiler-rt/this module to dump all the gathered coverage data to disk:

   llvm_gcda_start_file()
     llvm_gcda_emit_function()
     llvm_gcda_emit_arcs()
     llvm_gcda_emit_function()
     llvm_gcda_emit_arcs()
     [... repeats for each function ...]
   llvm_gcda_summary_info()
   llvm_gcda_end_file()

This design is much more stateless and unstructured than gcc's, and is
intended to run at process exit.  This forces us to keep some local
state about which module we're dealing with at the moment.  On the other
hand, it also means we don't depend as much on how LLVM represents
profiling data internally.

See LLVM's lib/Transforms/Instrumentation/GCOVProfiling.cpp for more
details on how this works, particularly GCOVProfiler::emitProfileArcs(),
GCOVProfiler::insertCounterWriteout(), and GCOVProfiler::insertFlush().

[akpm@linux-foundation.org: coding-style fixes]
Link: http://lkml.kernel.org/r/20190417225328.208129-1-trong@android.com
Signed-off-by: Greg Hackmann <ghackmann@android.com>
Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Tri Vo <trong@android.com>
Co-developed-by: Nick Desaulniers <ndesaulniers@google.com>
Co-developed-by: Tri Vo <trong@android.com>
Tested-by: Trilok Soni <tsoni@quicinc.com>
Tested-by: Prasad Sodagudi <psodagud@quicinc.com>
Tested-by: Tri Vo <trong@android.com>
Tested-by: Daniel Mentz <danielmentz@google.com>
Tested-by: Petri Gynther <pgynther@google.com>
Reviewed-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

gcov: docs: add a note on GCC vs Clang differences

Document some things of note to gcov users:
1. GCC gcov and Clang llvm-cov tools are not compatible.
2. The use of GCC vs Clang is transparent at build-time.

Also adjust the documentation to account for the removal of config symbol
CONFIG_GCOV_FORMAT_AUTODETECT by commit 6a61b70b43c9 ("gcov: remove
CONFIG_GCOV_FORMAT_AUTODETECT").

Link: http://lkml.kernel.org/r/20190318025411.98014-4-trong@android.com
Signed-off-by: Tri Vo <trong@android.com>
Reviewed-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Cc: Daniel Mentz <danielmentz@google.com>
Cc: Greg Hackmann <ghackmann@android.com>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Petri Gynther <pgynther@google.com>
Cc: Prasad Sodagudi <psodagud@quicinc.com>
Cc: Trilok Soni <tsoni@quicinc.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

gcov: clang: move common GCC code into gcc_base.c

Patch series "gcov: add Clang support", v4.

This patch (of 3):

base.c contains a few callbacks specific to GCC's gcov implementation.
Move these into their own module in preparation for Clang support.

Link: http://lkml.kernel.org/r/20190318025411.98014-2-trong@android.com
Signed-off-by: Greg Hackmann <ghackmann@android.com>
Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Tri Vo <trong@android.com>
Tested-by: Trilok Soni <tsoni@quicinc.com>
Tested-by: Prasad Sodagudi <psodagud@quicinc.com>
Tested-by: Tri Vo <trong@android.com>
Reviewed-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Cc: Daniel Mentz <danielmentz@google.com>
Cc: Petri Gynther <pgynther@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fs/eventfd.c: make eventfd_ida static

Fix sparse warning:

fs/eventfd.c:26:1: warning:
 symbol 'eventfd_ida' was not declared. Should it be static?

Link: http://lkml.kernel.org/r/20190413142348.34716-1-yuehaibing@huawei.com
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

eventfd: present id to userspace via fdinfo

Finding endpoints of an IPC channel is one of essential task to
understand how a user program works.  Procfs and netlink socket provide
enough hints to find endpoints for IPC channels like pipes, unix
sockets, and pseudo terminals.  However, there is no simple way to find
endpoints for an eventfd file from userland.  An inode number doesn't
hint.  Unlike pipe, all eventfd files share the same inode object.

To provide the way to find endpoints of an eventfd file, this patch adds
"eventfd-id" field to /proc/PID/fdinfo of eventfd as identifier.
Integers managed by an IDA are used as ids.

A tool like lsof can utilize the information to print endpoints.

Link: http://lkml.kernel.org/r/20190327181823.20222-1-yamato@redhat.com
Signed-off-by: Masatake YAMATO <yamato@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Kees Cook <keescook@chromium.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

kernel/pid.c: remove unneeded hash header file

Hash functions are not needed since idr is used now.  Let's remove hash
header file for cleanup.

Link: http://lkml.kernel.org/r/20190430053319.95913-1-scuttimmy@gmail.com
Signed-off-by: Timmy Li <scuttimmy@gmail.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: KJ Tsanaktsidis <ktsanaktsidis@zendesk.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

kernel/sysctl.c: fix proc_do_large_bitmap for large input buffers

Today, proc_do_large_bitmap() truncates a large write input buffer to
PAGE_SIZE - 1, which may result in misparsed numbers at the (truncated)
end of the buffer.  Further, it fails to notify the caller that the
buffer was truncated, so it doesn't get called iteratively to finish the
entire input buffer.

Tell the caller if there's more work to do by adding the skipped amount
back to left/*lenp before returning.

To fix the misparsing, reset the position if we have completely consumed
a truncated buffer (or if just one char is left, which may be a "-" in a
range), and ask the caller to come back for more.

Link: http://lkml.kernel.org/r/20190320222831.8243-7-mcgrof@kernel.org
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
Acked-by: Kees Cook <keescook@chromium.org>
Cc: Eric Sandeen <sandeen@sandeen.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

tools/testing/selftests/sysctl/sysctl.sh: add proc_do_large_bitmap() test case

The kernel has only two users of proc_do_large_bitmap(), the kernel CPU
watchdog, and the ip_local_reserved_ports.  Refer to watchdog_cpumask
and ip_local_reserved_ports in Documentation for further details on
these.  When you input a large buffer into these, when it is larger than
PAGE_SIZE- 1, the input data gets misparsed, and the user get
incorrectly informed that the desired input value was set.  This commit
implements a test which mimics and exploits that use case, it uses a
bitmap size, as in the watchdog case.  The bitmap is used to test the
bitmap proc handler, proc_do_large_bitmap().

The next commit fixes this issue.

[akpm@linux-foundation.org: move proc_do_large_bitmap() export to EOF]
[mcgrof@kernel.org: use new target description for backward compatibility]
[mcgrof@kernel.org: augment test number to 50, ran into issues with bash string comparisons when testing up to 50 cases.]
[mcgrof@kernel.org: introduce and use verify_diff_proc_file() to use diff]
[mcgrof@kernel.org: use mktemp for tmp file]
[mcgrof@kernel.org: merge shell test and C code]
[mcgrof@kernel.org: commit log love]
[mcgrof@kernel.org: export proc_do_large_bitmap() to allow for the test
[mcgrof@kernel.org: check for the return value when writing to the proc file]
Link: http://lkml.kernel.org/r/20190320222831.8243-6-mcgrof@kernel.org
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
Acked-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

tools/testing/selftests/sysctl/sysctl.sh: allow graceful use on older kernels

On old kernels older new test knobs implemented on the test_sysctl
module may not be available.  This is expected, and the selftests test
scripts should be able to run without failures on older kernels.

Generalize a solution so that we test for each required test target file
for each test by requiring each test description to annotate their
respective test target file.  If the target file does not exist, we skip
the test gracefully.

Link: http://lkml.kernel.org/r/20190320222831.8243-5-mcgrof@kernel.org
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
Acked-by: Kees Cook <keescook@chromium.org>
Cc: Eric Sandeen <sandeen@redhat.com>
Cc: Eric Sandeen <sandeen@sandeen.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

tools/testing/selftests/sysctl/sysctl.sh: ignore diff output on verify_diff_w()

When verify_diff_w() is used we care about the result, not the verbose
output, and although we use -q, that still gives us a chatty message
about if the files differ or not.  Since verify_diff_w() uses stdinput
the chatty message says whether or not "-" matches the target file, and
this just seems rather odd.  Better to just ignore that messsage all
together, what we really care about i sthe results, the return value and
we check for that.

Link: http://lkml.kernel.org/r/20190320222831.8243-4-mcgrof@kernel.org
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
Acked-by: Kees Cook <keescook@chromium.org>
Cc: Eric Sandeen <sandeen@redhat.com>
Cc: Eric Sandeen <sandeen@sandeen.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

tools/testing/selftests/sysctl/sysctl.sh: load module before testing for it

Currently the test script checks for the existence of the sysctl test
module's directory path prior to loading it.  We must first try to load
the module prior to checking for that path.  This fixes the order for
the load / test.

Link: http://lkml.kernel.org/r/20190320222831.8243-3-mcgrof@kernel.org
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
Acked-by: Kees Cook <keescook@chromium.org>
Cc: Eric Sandeen <sandeen@redhat.com>
Cc: Eric Sandeen <sandeen@sandeen.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

tools/testing/selftests/sysctl/sysctl.sh: remove superfluous test_reqs()

Patch series "sysctl: add pending proc_do_large_bitmap fix".

Eric sent a fix out for proc_do_large_bitmap() last month for when using
a large input buffer.  After patch review a test case for the issue was
built and submitted.  I noticed there were a few issues with the tests,
but instead of just asking Eric to address them I've taken care of them
and ammended the commit where necessary.  There's a few issues he
reported which I also address and fix in this series.

Since we *do* expect users of these scripts to also use them on older
kernels, I've also addressed not breaking calling the script for them,
and gives us an easy way to easily extend our tests cases for future
kernels as well.

Before anyone considers these for stable as minor fixes, I'd recommend
we also address the discrepancy on the read side of things: modify the
test script to use diff against the target file instead of using the
temp file.

This patch (of 6):

We already call test_reqs(), no need to call it twice.

Link: http://lkml.kernel.org/r/20190320222831.8243-2-mcgrof@kernel.org
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
Acked-by: Kees Cook <keescook@chromium.org>
Cc: Eric Sandeen <sandeen@redhat.com>
Cc: Eric Sandeen <sandeen@sandeen.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

sysctl: return -EINVAL if val violates minmax

Currently when userspace gives us a values that overflow e.g.  file-max
and other callers of __do_proc_doulongvec_minmax() we simply ignore the
new value and leave the current value untouched.

This can be problematic as it gives the illusion that the limit has
indeed be bumped when in fact it failed.  This commit makes sure to
return EINVAL when an overflow is detected.  Please note that this is a
userspace facing change.

Link: http://lkml.kernel.org/r/20190210203943.8227-4-christian@brauner.io
Signed-off-by: Christian Brauner <christian@brauner.io>
Acked-by: Luis Chamberlain <mcgrof@kernel.org>
Cc: Kees Cook <keescook@chromium.org>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Dominik Brodowski <linux@dominikbrodowski.net>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Joe Lawrence <joe.lawrence@redhat.com>
Cc: Waiman Long <longman@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

kernel/sysctl.c: switch to bitmap_zalloc()

Switch to bitmap_zalloc() to show clearly what we are allocating.
Besides that it returns pointer of bitmap type instead of opaque void *.

Link: http://lkml.kernel.org/r/20190304094037.57756-1-andriy.shevchenko@linux.intel.com
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Acked-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Luis Chamberlain <mcgrof@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

rapidio: fix a NULL pointer dereference when create_workqueue() fails

In case create_workqueue fails, the fix releases resources and returns
-ENOMEM to avoid NULL pointer dereference.

Signed-off-by: Kangjie Lu <kjlu@umn.edu>
Acked-by: Alexandre Bounine <alex.bou9@gmail.com>
Cc: Matt Porter <mporter@kernel.crashing.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

include/linux/cpumask.h: fix double string traverse in cpumask_parse

cpumask_parse() finds first occurrence of either or strchr() and
strlen().  We can do it better with a single call of strchrnul().

[akpm@linux-foundation.org: remove unneeded cast]
Link: http://lkml.kernel.org/r/20190409204208.12190-1-ynorov@marvell.com
Signed-off-by: Yury Norov <ynorov@marvell.com>
Acked-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

exec selftests: test ->recursion_depth

Test that trivially recursing script onto itself doesn't work.

Note: this is different test from ELOOP tests in execveat.c Those test
that execveat(2) doesn't follow symlinks when told to do so.

Link: http://lkml.kernel.org/r/20190423192720.GA21433@avx2
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

exec: move struct linux_binprm::buf

struct linux_binprm::buf is the first field and it is exactly 128 bytes
in size.  It means that on x86_64 all accesses to other fields will go
though [r64 + disp32] addressing mode which is 3 bytes bloatier than
[r64 + disp8] addressing mode.  Given that accesses to other fields
outnumber accesses to ->buf, move it down.

Space savings (x86_64 defconfig):
more on distro configs because LSMs actively dereference "bprm"
but do not care about first 128 bytes of the executable itself.

add/remove: 0/0 grow/shrink: 0/24 up/down: 0/-492 (-492)
Function                                     old     new   delta
selinux_bprm_committing_creds                552     549      -3
finalize_exec                                 94      91      -3
__audit_log_bprm_fcaps                       283     280      -3
__audit_bprm                                  39      36      -3
perf_trace_sched_process_exec                347     341      -6
install_exec_creds                           105      99      -6
cap_bprm_set_creds.cold                       60      54      -6
would_dump                                   137     128      -9
load_script                                  637     628      -9
bprm_change_interp                            61      52      -9
trace_event_raw_event_sched_process_exec     260     250     -10
search_binary_handler                        255     240     -15
remove_arg_zero                              295     277     -18
free_bprm                                    119     101     -18
prepare_binprm                               379     360     -19
setup_new_exec                               336     315     -21
flush_old_exec                              1638    1617     -21
copy_strings.isra                            746     724     -22
setup_arg_pages                              559     530     -29
load_misc_binary                            1151    1118     -33
selinux_bprm_set_creds                       792     753     -39
load_elf_binary                            11111   11072     -39
cap_bprm_set_creds                          1496    1454     -42
__do_execve_file.isra                       2395    2286    -109

Link: http://lkml.kernel.org/r/20190421165025.GA26843@avx2
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fs/exec.c: move ->recursion_depth out of critical sections

->recursion_depth is changed only by current, therefore decrementing can
be done without taking any locks.

Link: http://lkml.kernel.org/r/20190417213150.GA26474@avx2
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

kernel/signal.c: annotate implicit fall through

There is a plan to build the kernel with -Wimplicit-fallthrough and this
place in the code produced a warning (W=1).

This commit remove the following warning:

  kernel/signal.c:795:13: warning: this statement may fall through [-Wimplicit-fallthrough=]

Link: http://lkml.kernel.org/r/20190114203505.17875-1-malat@debian.org
Signed-off-by: Mathieu Malaterre <malat@debian.org>
Acked-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fs/fat/file.c: issue flush after the writeback of FAT

fsync() needs to make sure the data & meta-data of file are persistent
after the return of fsync(), even when a power-failure occurs later.  In
the case of fat-fs, the FAT belongs to the meta-data of file, so we need
to issue a flush after the writeback of FAT instead before.

Also bail out early when any stage of fsync fails.

Link: http://lkml.kernel.org/r/20190409030158.136316-1-houtao1@huawei.com
Signed-off-by: Hou Tao <houtao1@huawei.com>
Acked-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

reiserfs: add comment to explain endianness issue in xattr_hash

csum_partial() gives different results for little-endian and big-endian
hosts.  This causes images created on little-endian hosts and mounted on
big endian hosts to see csum mismatches.  This causes an endianness bug.
Sparse gives a warning as csum_partial returns a restricted integer type
__wsum_t and xattr_hash expects __u32.  This warning acts as a reminder
for this bug and should not be suppressed.

This comment aims to convey these endianness issues.

[akpm@linux-foundation.org: coding-style fixes]
Link: http://lkml.kernel.org/r/20190423161831.GA15387@bharath12345-Inspiron-5559
Signed-off-by: Bharath Vedartham <linux.bhar@gmail.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Jann Horn <jannh@google.com>
Cc: Jeff Mahoney <jeffm@suse.com>
Cc: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

autofs: add description of ignore pseudo mount option

Add a description of the "ignore" pseudo mount option that can be used
to provide a generic indicator to applications that the mount entry
should be ignored when displaying mount information.

Link: http://lkml.kernel.org/r/155287084617.12593.812733161112154904.stgit@pluto.themaw.net
Signed-off-by: Ian Kent <raven@themaw.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

autofs: update mount control expire desription with AUTOFS_EXP_FORCED

Describe AUTOFS_EXP_FORCED in addition to AUTOFS_EXP_IMMEDIATE in the
description of the AUTOFS_DEV_IOCTL_EXPIRE_CMD ioctl.

Link: http://lkml.kernel.org/r/155287084078.12593.15000931045413195778.stgit@pluto.themaw.net
Signed-off-by: Ian Kent <raven@themaw.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

autofs: update AUTOFS_EXP_LEAVES description

Update the description of AUTOFS_EXP_LEAVES to cover its possible future
use with amd format mount maps.

Link: http://lkml.kernel.org/r/155287083538.12593.18163159677020718048.stgit@pluto.themaw.net
Signed-off-by: Ian Kent <raven@themaw.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

autofs: update autofs.txt for strictexpire mount option

A "strictexpire" mount option has been added to the autofs file system.

It is meant to be used in cases where a GUI continually accesses or an
application frquently scans an automount directory tree causing an
accumulation of otherwise unused mounts.

Link: http://lkml.kernel.org/r/155287083000.12593.2722713092537666885.stgit@pluto.themaw.net
Signed-off-by: Ian Kent <raven@themaw.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

autofs: fix some word usage oddities in autofs.txt

Alter a few word usages in Documentation/filesystems/autofs.txt and
correct some spelling mistakes.

Link: http://lkml.kernel.org/r/155287082394.12593.6506084453911662450.stgit@pluto.themaw.net
Signed-off-by: Ian Kent <raven@themaw.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

net: replace CONFIG_DEBUG_KERNEL with CONFIG_DEBUG_MISC

CONFIG_DEBUG_KERNEL should not impact code generation.  Use the newly
defined CONFIG_DEBUG_MISC instead to keep the current code.

Link: http://lkml.kernel.org/r/20190413224438.10802-6-okaya@kernel.org
Signed-off-by: Sinan Kaya <okaya@kernel.org>
Acked-by: Florian Westphal <fw@strlen.de>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
Reviewed-by: Kees Cook <keescook@chromium.org>
Cc: Florian Westphal <fw@strlen.de>
Cc: Pablo Neira Ayuso <pablo@netfilter.org>
Cc: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Anders Roxell <anders.roxell@linaro.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Christophe Leroy <christophe.leroy@c-s.fr>
Cc: Chris Zankel <chris@zankel.net>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: James Hogan <jhogan@kernel.org>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Paul Burton <paul.burton@mips.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Thomas Bogendoerfer <tbogendoerfer@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

xtensa: replace CONFIG_DEBUG_KERNEL with CONFIG_DEBUG_MISC

CONFIG_DEBUG_KERNEL should not impact code generation.  Use the newly
defined CONFIG_DEBUG_MISC instead to keep the current code.

Link: http://lkml.kernel.org/r/20190413224438.10802-5-okaya@kernel.org
Signed-off-by: Sinan Kaya <okaya@kernel.org>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
Reviewed-by: Kees Cook <keescook@chromium.org>
Acked-by: Max Filippov <jcmvbkbc@gmail.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: Anders Roxell <anders.roxell@linaro.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Christophe Leroy <christophe.leroy@c-s.fr>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Florian Westphal <fw@strlen.de>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: James Hogan <jhogan@kernel.org>
Cc: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Pablo Neira Ayuso <pablo@netfilter.org>
Cc: Paul Burton <paul.burton@mips.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Thomas Bogendoerfer <tbogendoerfer@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

powerpc: replace CONFIG_DEBUG_KERNEL with CONFIG_DEBUG_MISC

CONFIG_DEBUG_KERNEL should not impact code generation.  Use the newly
defined CONFIG_DEBUG_MISC instead to keep the current code.

Link: http://lkml.kernel.org/r/20190413224438.10802-3-okaya@kernel.org
Signed-off-by: Sinan Kaya <okaya@kernel.org>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
Reviewed-by: Kees Cook <keescook@chromium.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Christophe Leroy <christophe.leroy@c-s.fr>
Cc: Anders Roxell <anders.roxell@linaro.org>
Cc: Chris Zankel <chris@zankel.net>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Florian Westphal <fw@strlen.de>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: James Hogan <jhogan@kernel.org>
Cc: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Pablo Neira Ayuso <pablo@netfilter.org>
Cc: Paul Burton <paul.burton@mips.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Thomas Bogendoerfer <tbogendoerfer@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

init: introduce DEBUG_MISC option

Patch series "init: Do not select DEBUG_KERNEL by default", v5.

CONFIG_DEBUG_KERNEL has been designed to just enable Kconfig options.
Kernel code generatoin should not depend on CONFIG_DEBUG_KERNEL.

Proposed alternative plan: let's add a new symbol, something like
DEBUG_MISC ("Miscellaneous debug code that should be under a more
specific debug option but isn't"), make it depend on DEBUG_KERNEL and be
"default DEBUG_KERNEL" but allow itself to be turned off, and then
mechanically change the small handful of "#ifdef CONFIG_DEBUG_KERNEL" to
"#ifdef CONFIG_DEBUG_MISC".

This patch (of 5):

Introduce DEBUG_MISC ("Miscellaneous debug code that should be under a
more specific debug option but isn't"), make it depend on DEBUG_KERNEL
and be "default DEBUG_KERNEL" but allow itself to be turned off, and
then mechanically change the small handful of "#ifdef
CONFIG_DEBUG_KERNEL" to "#ifdef CONFIG_DEBUG_MISC".

Link: http://lkml.kernel.org/r/20190413224438.10802-2-okaya@kernel.org
Signed-off-by: Sinan Kaya <okaya@kernel.org>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
Reviewed-by: Kees Cook <keescook@chromium.org>
Cc: Anders Roxell <anders.roxell@linaro.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Christophe Leroy <christophe.leroy@c-s.fr>
Cc: Chris Zankel <chris@zankel.net>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Florian Westphal <fw@strlen.de>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: James Hogan <jhogan@kernel.org>
Cc: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Pablo Neira Ayuso <pablo@netfilter.org>
Cc: Paul Burton <paul.burton@mips.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Thomas Bogendoerfer <tbogendoerfer@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

binfmt_elf: move brk out of mmap when doing direct loader exec

Commmit eab09532d400 ("binfmt_elf: use ELF_ET_DYN_BASE only for PIE"),
made changes in the rare case when the ELF loader was directly invoked
(e.g to set a non-inheritable LD_LIBRARY_PATH, testing new versions of
the loader), by moving into the mmap region to avoid both ET_EXEC and
PIE binaries.  This had the effect of also moving the brk region into
mmap, which could lead to the stack and brk being arbitrarily close to
each other.  An unlucky process wouldn't get its requested stack size
and stack allocations could end up scribbling on the heap.

This is illustrated here.  In the case of using the loader directly, brk
(so helpfully identified as "[heap]") is allocated with the _loader_ not
the binary.  For example, with ASLR entirely disabled, you can see this
more clearly:

$ /bin/cat /proc/self/maps
555555554000-55555555c000 r-xp 00000000 ... /bin/cat
55555575b000-55555575c000 r--p 00007000 ... /bin/cat
55555575c000-55555575d000 rw-p 00008000 ... /bin/cat
55555575d000-55555577e000 rw-p 00000000 ... [heap]
...
7ffff7ff7000-7ffff7ffa000 r--p 00000000 ... [vvar]
7ffff7ffa000-7ffff7ffc000 r-xp 00000000 ... [vdso]
7ffff7ffc000-7ffff7ffd000 r--p 00027000 ... /lib/x86_64-linux-gnu/ld-2.27.so
7ffff7ffd000-7ffff7ffe000 rw-p 00028000 ... /lib/x86_64-linux-gnu/ld-2.27.so
7ffff7ffe000-7ffff7fff000 rw-p 00000000 ...
7ffffffde000-7ffffffff000 rw-p 00000000 ... [stack]

$ /lib/x86_64-linux-gnu/ld-2.27.so /bin/cat /proc/self/maps
...
7ffff7bcc000-7ffff7bd4000 r-xp 00000000 ... /bin/cat
7ffff7bd4000-7ffff7dd3000 ---p 00008000 ... /bin/cat
7ffff7dd3000-7ffff7dd4000 r--p 00007000 ... /bin/cat
7ffff7dd4000-7ffff7dd5000 rw-p 00008000 ... /bin/cat
7ffff7dd5000-7ffff7dfc000 r-xp 00000000 ... /lib/x86_64-linux-gnu/ld-2.27.so
7ffff7fb2000-7ffff7fd6000 rw-p 00000000 ...
7ffff7ff7000-7ffff7ffa000 r--p 00000000 ... [vvar]
7ffff7ffa000-7ffff7ffc000 r-xp 00000000 ... [vdso]
7ffff7ffc000-7ffff7ffd000 r--p 00027000 ... /lib/x86_64-linux-gnu/ld-2.27.so
7ffff7ffd000-7ffff7ffe000 rw-p 00028000 ... /lib/x86_64-linux-gnu/ld-2.27.so
7ffff7ffe000-7ffff8020000 rw-p 00000000 ... [heap]
7ffffffde000-7ffffffff000 rw-p 00000000 ... [stack]

The solution is to move brk out of mmap and into ELF_ET_DYN_BASE since
nothing is there in the direct loader case (and ET_EXEC is still far
away at 0x400000).  Anything that ran before should still work (i.e.
the ultimately-launched binary already had the brk very far from its
text, so this should be no different from a COMPAT_BRK standpoint).  The
only risk I see here is that if someone started to suddenly depend on
the entire memory space lower than the mmap region being available when
launching binaries via a direct loader execs which seems highly
unlikely, I'd hope: this would mean a binary would _not_ work when
exec()ed normally.

(Note that this is only done under CONFIG_ARCH_HAS_ELF_RANDOMIZATION
when randomization is turned on.)

Link: http://lkml.kernel.org/r/20190422225727.GA21011@beast
Link: https://lkml.kernel.org/r/CAGXu5jJ5sj3emOT2QPxQkNQk0qbU6zEfu9=Omfhx_p0nCKPSjA@mail.gmail.com
Fixes: eab09532d400 ("binfmt_elf: use ELF_ET_DYN_BASE only for PIE")
Signed-off-by: Kees Cook <keescook@chromium.org>
Reported-by: Ali Saidi <alisaidi@amazon.com>
Cc: Ali Saidi <alisaidi@amazon.com>
Cc: Guenter Roeck <linux@roeck-us.net>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Jann Horn <jannh@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

elf: init pt_regs pointer later

Get "current_pt_regs" pointer right before usage.

Space savings on x86_64:

add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-180 (-180)
Function                           old     new   delta
load_elf_binary                   5806    5626    -180 !!!

Looks like the compiler doesn't know that "current_pt_regs" is stable
pointer (because it doesn't know ->stack isn't) even though it knows
that "current" is stable pointer.  So it saves it in the very beginning
and then tries to carry it through a lot of code.

Here is what happens here:

load_elf_binary()
...
mov rax,QWORD PTR gs:0x14c00
mov r13,QWORD PTR [rax+0x18] r13 = current->stack
call kmem_cache_alloc # first kmalloc

[980 bytes later!]

# let's spill that sucker because we need a register
# for "load_bias" calculations at
#
# if (interpreter) {
# load_bias = ELF_ET_DYN_BASE;
# if (current->flags & PF_RANDOMIZE)
# load_bias += arch_mmap_rnd();
# elf_flags |= elf_fixed;
# }
mov QWORD PTR [rsp+0x68],r13

If this is not _the_ root cause it is still eeeeh.

After the patch things become much simpler:

mov rax, QWORD PTR gs:0x14c00 # current
mov rdx, QWORD PTR [rax+0x18] # current->stack
movq [rdx+0x3fb8], 0 # fill pt_regs
...
call finalize_exec

Link: http://lkml.kernel.org/r/20190419200343.GA19788@avx2
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Tested-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fs/binfmt_elf.c: extract PROT_* calculations

There are two places where mapping protections are calculated: one for
executable, another one for interpreter -- take them out.

ELF read and execute permissions are interchanged with Linux PROT_READ
and PROT_EXEC, microoptimizations are welcome!

Link: http://lkml.kernel.org/r/20190417213413.GB26474@avx2
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fs//binfmt_elf.c: move variables initialization closer to their usage

Link: http://lkml.kernel.org/r/20190416202002.GB24304@avx2
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fs/binfmt_elf.c: save 1 indent level

Rewrite

for (...) {
if (->p_type == PT_INTERP) {
...
break;
}
}

loop into

for (...) {
if (->p_type != PT_INTERP)
continue;
...
break;
}

Link: http://lkml.kernel.org/r/20190416201906.GA24304@avx2
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fs/binfmt_elf.c: delete trailing "return;" in functions returning "void"

Link: http://lkml.kernel.org/r/20190314205042.GE18143@avx2
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fs/binfmt_elf.c: free PT_INTERP filename ASAP

There is no reason for PT_INTERP filename to linger till the end of the
whole loading process.

Link: http://lkml.kernel.org/r/20190314204953.GD18143@avx2
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Nikitas Angelinas <nikitas.angelinas@gmail.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Mukesh Ojha <mojha@codeaurora.org>
[nikitas.angelinas@gmail.com: fix GPF when dereferencing invalid interpreter]
Link: http://lkml.kernel.org/r/20190330140032.GA1527@vostro
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fs/binfmt_elf.c: make scope of "pos" variable smaller

Link: http://lkml.kernel.org/r/20190314204707.GC18143@avx2
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fs/binfmt_elf.c: remove unneeded initialization of mm->start_stack

As pointed out by zoujc@lenovo.com, setup_arg_pages() already
initialized current->mm->start_stack.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=202881
Reported-by: <zoujc@lenovo.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

lib/test_vmalloc.c:test_func(): eliminate local `ret'

Local 'ret' is unneeded and was poorly named: the variable `ret'
generally means the "the value which this function will return".

Cc: Roman Gushchin <guro@fb.com>
Cc: Uladzislau Rezki <urezki@gmail.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Thomas Garnier <thgarnie@google.com>
Cc: Oleksiy Avramchenko <oleksiy.avramchenko@sonymobile.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Joel Fernandes <joelaf@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

include/linux/bitops.h: sanitize rotate primitives

The ror32 implementation (word >> shift) | (word << (32 - shift) has
undefined behaviour if shift is outside the [1, 31] range.  Similarly
for the 64 bit variants.  Most callers pass a compile-time constant
(naturally in that range), but there's an UBSAN report that these may
actually be called with a shift count of 0.

Instead of special-casing that, we can make them DTRT for all values of
shift while also avoiding UB.  For some reason, this was already partly
done for rol32 (which was well-defined for [0, 31]).  gcc 8 recognizes
these patterns as rotates, so for example

  __u32 rol32(__u32 word, unsigned int shift)
  {
return (word << (shift & 31)) | (word >> ((-shift) & 31));
  }

compiles to

0000000000000020 <rol32>:
  20:   89 f8                   mov    %edi,%eax
  22:   89 f1                   mov    %esi,%ecx
  24:   d3 c0                   rol    %cl,%eax
  26:   c3                      retq

Older compilers unfortunately do not do as well, but this only affects
the small minority of users that don't pass constants.

Due to integer promotions, ro[lr]8 were already well-defined for shifts
in [0, 8], and ro[lr]16 were mostly well-defined for shifts in [0, 16]
(only mostly - u16 gets promoted to _signed_ int, so if bit 15 is set,
word << 16 is undefined).  For consistency, update those as well.

Link: http://lkml.kernel.org/r/20190410211906.2190-1-linux@rasmusvillemoes.dk
Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Reported-by: Ido Schimmel <idosch@mellanox.com>
Tested-by: Ido Schimmel <idosch@mellanox.com>
Reviewed-by: Will Deacon <will.deacon@arm.com>
Cc: Vadim Pasternak <vadimp@mellanox.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Jacek Anaszewski <jacek.anaszewski@gmail.com>
Cc: Pavel Machek <pavel@ucw.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

lib/test_bitmap: add tests for bitmap_parselist_user()

Propagate existing bitmap_parselist() tests to bitmap_parselist_user().

Link: http://lkml.kernel.org/r/20190405173211.11373-6-ynorov@marvell.com
Signed-off-by: Yury Norov <ynorov@marvell.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Kees Cook <keescook@chromium.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mike Travis <travis@sgi.com>
Cc: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

lib/test_bitmap: add testcases for bitmap_parselist()

Add tests for non-number character, empty regions, integer overflow.

[ynorov@marvell.com: v5]
Link: http://lkml.kernel.org/r/20190416063801.20134-5-ynorov@marvell.com
Link: http://lkml.kernel.org/r/20190405173211.11373-5-ynorov@marvell.com
Signed-off-by: Yury Norov <ynorov@marvell.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Kees Cook <keescook@chromium.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mike Travis <travis@sgi.com>
Cc: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

lib/test_bitmap: switch test_bitmap_parselist to ktime_get()

test_bitmap_parselist currently uses get_cycles which is not implemented
on some platforms, so use ktime_get() instead.

Link: http://lkml.kernel.org/r/20190405173211.11373-4-ynorov@marvell.com
Signed-off-by: Yury Norov <ynorov@marvell.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Kees Cook <keescook@chromium.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mike Travis <travis@sgi.com>
Cc: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

lib: rework bitmap_parselist

Remove __bitmap_parselist helper and split the function to logical
parts.

[ynorov@marvell.com: v5]
Link: http://lkml.kernel.org/r/20190416063801.20134-3-ynorov@marvell.com
Link: http://lkml.kernel.org/r/20190405173211.11373-3-ynorov@marvell.com
Signed-off-by: Yury Norov <ynorov@marvell.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Kees Cook <keescook@chromium.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mike Travis <travis@sgi.com>
Cc: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

lib: make bitmap_parselist_user() a wrapper on bitmap_parselist()

Patch series "lib: rework bitmap_parselist and tests", v5.

bitmap_parselist has been evolved from a pretty simple idea for long and
now lacks for refactoring.  It is not structured, has nested loops and a
set of opaque-named variables.

Things are more complicated because bitmap_parselist() is a part of user
interface, and its behavior should not change.

In this patchset
 - bitmap_parselist_user() made a wrapper on bitmap_parselist();
 - bitmap_parselist() reworked (patch 2);
 - time measurement in test_bitmap_parselist switched to ktime_get
   (patch 3);
 - new tests introduced (patch 4), and
 - bitmap_parselist_user() testing enabled with the same testset as
   bitmap_parselist() (patch 5).

This patch (of 5):

Currently we parse user data byte after byte which leads to
overcomplification of parsing algorithm.  The only user of
bitmap_parselist_user() is not performance-critical, and so we can
duplicate user data to kernel buffer and simply call bitmap_parselist().
This rework lets us unify and simplify bitmap_parselist() and
bitmap_parselist_user(), which is done in the following patch.

Link: http://lkml.kernel.org/r/20190405173211.11373-2-ynorov@marvell.com
Signed-off-by: Yury Norov <ynorov@marvell.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Kees Cook <keescook@chromium.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: Mike Travis <travis@sgi.com>
Cc: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

lib/math: move int_pow() from pwm_bl.c for wider use

The integer exponentiation is used in few places and might be used in
the future by other call sites.  Move it to wider use.

Link: http://lkml.kernel.org/r/20190323172531.80025-2-andriy.shevchenko@linux.intel.com
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: Daniel Thompson <daniel.thompson@linaro.org>
Cc: Lee Jones <lee.jones@linaro.org>
Cc: Ray Jui <rjui@broadcom.com>
Cc: Thierry Reding <thierry.reding@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

lib: Move mathematic helpers to separate folder

For better maintenance and expansion move the mathematic helpers to the
separate folder.

No functional change intended.

Note, the int_sqrt() is not used as a part of lib, so, moved to regular
obj.

Link: http://lkml.kernel.org/r/20190323172531.80025-1-andriy.shevchenko@linux.intel.com
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Thierry Reding <thierry.reding@gmail.com>
Cc: Lee Jones <lee.jones@linaro.org>
Cc: Daniel Thompson <daniel.thompson@linaro.org>
Cc: Ray Jui <rjui@broadcom.com>
[mchehab+samsung@kernel.org: fix broken doc references for div64.c and gcd.c]
Link: http://lkml.kernel.org/r/734f49bae5d4052b3c25691dfefad59bea2e5843.1555580999.git.mchehab+samsung@kernel.org
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

lib/list_sort: optimize number of calls to comparison function

CONFIG_RETPOLINE has severely degraded indirect function call
performance, so it's worth putting some effort into reducing the number
of times cmp() is called.

This patch avoids badly unbalanced merges on unlucky input sizes.  It
slightly increases the code size, but saves an average of 0.2*n calls to
cmp().

x86-64 code size 739 -> 803 bytes (+64)

Unfortunately, there's not a lot of low-hanging fruit in a merge sort;
it already performs only n*log2(n) - K*n + O(1) compares.  The leading
coefficient is already at the theoretical limit (log2(n!) corresponds to
K=1.4427), so we're fighting over the linear term, and the best
mergesort can do is K=1.2645, achieved when n is a power of 2.

The differences between mergesort variants appear when n is *not* a
power of 2; K is a function of the fractional part of log2(n).  Top-down
mergesort does best of all, achieving a minimum K=1.2408, and an average
(over all sizes) K=1.248.  However, that requires knowing the number of
entries to be sorted ahead of time, and making a full pass over the
input to count it conflicts with a second performance goal, which is
cache blocking.

Obviously, we have to read the entire list into L1 cache at some point,
and performance is best if it fits.  But if it doesn't fit, each full
pass over the input causes a cache miss per element, which is
undesirable.

While textbooks explain bottom-up mergesort as a succession of merging
passes, practical implementations do merging in depth-first order: as
soon as two lists of the same size are available, they are merged.  This
allows as many merge passes as possible to fit into L1; only the final
few merges force cache misses.

This cache-friendly depth-first merge order depends on us merging the
beginning of the input as much as possible before we've even seen the
end of the input (and thus know its size).

The simple eager merge pattern causes bad performance when n is just
over a power of 2.  If n=1028, the final merge is between 1024- and
4-element lists, which is wasteful of comparisons.  (This is actually
worse on average than n=1025, because a 1204:1 merge will, on average,
end after 512 compares, while 1024:4 will walk 4/5 of the list.)

Because of this, bottom-up mergesort achieves K < 0.5 for such sizes,
and has an average (over all sizes) K of around 1.  (My experiments show
K=1.01, while theory predicts K=0.965.)

There are "worst-case optimal" variants of bottom-up mergesort which
avoid this bad performance, but the algorithms given in the literature,
such as queue-mergesort and boustrodephonic mergesort, depend on the
breadth-first multi-pass structure that we are trying to avoid.

This implementation is as eager as possible while ensuring that all
merge passes are at worst 1:2 unbalanced.  This achieves the same
average K=1.207 as queue-mergesort, which is 0.2*n better then
bottom-up, and only 0.04*n behind top-down mergesort.

Specifically, defers merging two lists of size 2^k until it is known
that there are 2^k additional inputs following.  This ensures that the
final uneven merges triggered by reaching the end of the input will be
at worst 2:1.  This will avoid cache misses as long as 3*2^k elements
fit into the cache.

(I confess to being more than a little bit proud of how clean this code
turned out.  It took a lot of thinking, but the resultant inner loop is
very simple and efficient.)

Refs:
  Bottom-up Mergesort: A Detailed Analysis
  Wolfgang Panny, Helmut Prodinger
  Algorithmica 14(4):340--354, October 1995
  https://doi.org/10.1007/BF01294131
  https://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.6.5260

  The cost distribution of queue-mergesort, optimal mergesorts, and
  power-of-two rules
  Wei-Mei Chen, Hsien-Kuei Hwang, Gen-Huey Chen
  Journal of Algorithms 30(2); Pages 423--448, February 1999
  https://doi.org/10.1006/jagm.1998.0986
  https://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.4.5380

  Queue-Mergesort
  Mordecai J. Golin, Robert Sedgewick
  Information Processing Letters, 48(5):253--259, 10 December 1993
  https://doi.org/10.1016/0020-0190(93)90088-q
  https://sci-hub.tw/10.1016/0020-0190(93)90088-Q

Feedback from Rasmus Villemoes <linux@rasmusvillemoes.dk>.

Link: http://lkml.kernel.org/r/fd560853cc4dca0d0f02184ffa888b4c1be89abc.1552704200.git.lkml@sdf.org
Signed-off-by: George Spelvin <lkml@sdf.org>
Acked-by: Andrey Abramov <st5pub@yandex.ru>
Acked-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: Daniel Wagner <daniel.wagner@siemens.com>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Don Mullis <don.mullis@gmail.com>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

lib/list_sort: simplify and remove MAX_LIST_LENGTH_BITS

Rather than a fixed-size array of pending sorted runs, use the ->prev
links to keep track of things.  This reduces stack usage, eliminates
some ugly overflow handling, and reduces the code size.

Also:
* merge() no longer needs to handle NULL inputs, so simplify.
* The same applies to merge_and_restore_back_links(), which is renamed
  to the less ponderous merge_final().  (It's a static helper function,
  so we don't need a super-descriptive name; comments will do.)
* Document the actual return value requirements on the (*cmp)()
  function; some callers are already using this feature.

x86-64 code size 1086 -> 739 bytes (-347)

(Yes, I see checkpatch complaining about no space after comma in
"__attribute__((nonnull(2,3,4,5)))".  Checkpatch is wrong.)

Feedback from Rasmus Villemoes, Andy Shevchenko and Geert Uytterhoeven.

[akpm@linux-foundation.org: remove __pure usage due to mysterious warning]
Link: http://lkml.kernel.org/r/f63c410e0ff76009c9b58e01027e751ff7fdb749.1552704200.git.lkml@sdf.org
Signed-off-by: George Spelvin <lkml@sdf.org>
Acked-by: Andrey Abramov <st5pub@yandex.ru>
Acked-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Daniel Wagner <daniel.wagner@siemens.com>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Don Mullis <don.mullis@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

lib/sort: avoid indirect calls to built-in swap

Similar to what's being done in the net code, this takes advantage of
the fact that most invocations use only a few common swap functions, and
replaces indirect calls to them with (highly predictable) conditional
branches.  (The downside, of course, is that if you *do* use a custom
swap function, there are a few extra predicted branches on the code
path.)

This actually *shrinks* the x86-64 code, because it inlines the various
swap functions inside do_swap, eliding function prologues & epilogues.

x86-64 code size 767 -> 703 bytes (-64)

Link: http://lkml.kernel.org/r/d10c5d4b393a1847f32f5b26f4bbaa2857140e1e.1552704200.git.lkml@sdf.org
Signed-off-by: George Spelvin <lkml@sdf.org>
Acked-by: Andrey Abramov <st5pub@yandex.ru>
Acked-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: Daniel Wagner <daniel.wagner@siemens.com>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Don Mullis <don.mullis@gmail.com>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

lib/sort: use more efficient bottom-up heapsort variant

This uses fewer comparisons than the previous code (approaching half as
many for large random inputs), but produces identical results; it
actually performs the exact same series of swap operations.

Specifically, it reduces the average number of compares from
  2*n*log2(n) - 3*n + o(n)
to
    n*log2(n) + 0.37*n + o(n).

This is still 1.63*n worse than glibc qsort() which manages n*log2(n) -
1.26*n, but at least the leading coefficient is correct.

Standard heapsort, when sifting down, performs two comparisons per
level: one to find the greater child, and a second to see if the current
node should be exchanged with that child.

Bottom-up heapsort observes that it's better to postpone the second
comparison and search for the leaf where -infinity would be sent to,
then search back *up* for the current node's destination.

Since sifting down usually proceeds to the leaf level (that's where half
the nodes are), this does O(1) second comparisons rather than log2(n).
That saves a lot of (expensive since Spectre) indirect function calls.

The one time it's worse than the previous code is if there are large
numbers of duplicate keys, when the top-down algorithm is O(n) and
bottom-up is O(n log n).  For distinct keys, it's provably always
better, doing 1.5*n*log2(n) + O(n) in the worst case.

(The code is not significantly more complex.  This patch also merges the
heap-building and -extracting sift-down loops, resulting in a net code
size savings.)

x86-64 code size 885 -> 767 bytes (-118)

(I see the checkpatch complaint about "else if (n -= size)".  The
alternative is significantly uglier.)

Link: http://lkml.kernel.org/r/2de8348635a1a421a72620677898c7fd5bd4b19d.1552704200.git.lkml@sdf.org
Signed-off-by: George Spelvin <lkml@sdf.org>
Acked-by: Andrey Abramov <st5pub@yandex.ru>
Acked-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: Daniel Wagner <daniel.wagner@siemens.com>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Don Mullis <don.mullis@gmail.com>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

lib/sort: make swap functions more generic

Patch series "lib/sort & lib/list_sort: faster and smaller", v2.

Because CONFIG_RETPOLINE has made indirect calls much more expensive, I
thought I'd try to reduce the number made by the library sort functions.

The first three patches apply to lib/sort.c.

Patch #1 is a simple optimization.  The built-in swap has special cases
for aligned 4- and 8-byte objects.  But those are almost never used;
most calls to sort() work on larger structures, which fall back to the
byte-at-a-time loop.  This generalizes them to aligned *multiples* of 4
and 8 bytes.  (If nothing else, it saves an awful lot of energy by not
thrashing the store buffers as much.)

Patch #2 grabs a juicy piece of low-hanging fruit.  I agree that nice
simple solid heapsort is preferable to more complex algorithms (sorry,
Andrey), but it's possible to implement heapsort with far fewer
comparisons (50% asymptotically, 25-40% reduction for realistic sizes)
than the way it's been done up to now.  And with some care, the code
ends up smaller, as well.  This is the "big win" patch.

Patch #3 adds the same sort of indirect call bypass that has been added
to the net code of late.  The great majority of the callers use the
builtin swap functions, so replace the indirect call to sort_func with a
(highly preditable) series of if() statements.  Rather surprisingly,
this decreased code size, as the swap functions were inlined and their
prologue & epilogue code eliminated.

lib/list_sort.c is a bit trickier, as merge sort is already close to
optimal, and we don't want to introduce triumphs of theory over
practicality like the Ford-Johnson merge-insertion sort.

Patch #4, without changing the algorithm, chops 32% off the code size
and removes the part[MAX_LIST_LENGTH+1] pointer array (and the
corresponding upper limit on efficiently sortable input size).

Patch #5 improves the algorithm.  The previous code is already optimal
for power-of-two (or slightly smaller) size inputs, but when the input
size is just over a power of 2, there's a very unbalanced final merge.

There are, in the literature, several algorithms which solve this, but
they all depend on the "breadth-first" merge order which was replaced by
commit 835cc0c8477f with a more cache-friendly "depth-first" order.
Some hard thinking came up with a depth-first algorithm which defers
merges as little as possible while avoiding bad merges.  This saves
0.2*n compares, averaged over all sizes.

The code size increase is minimal (64 bytes on x86-64, reducing the net
savings to 26%), but the comments expanded significantly to document the
clever algorithm.

TESTING NOTES: I have some ugly user-space benchmarking code which I
used for testing before moving this code into the kernel.  Shout if you
want a copy.

I'm running this code right now, with CONFIG_TEST_SORT and
CONFIG_TEST_LIST_SORT, but I confess I haven't rebooted since the last
round of minor edits to quell checkpatch.  I figure there will be at
least one round of comments and final testing.

This patch (of 5):

Rather than having special-case swap functions for 4- and 8-byte
objects, special-case aligned multiples of 4 or 8 bytes.  This speeds up
most users of sort() by avoiding fallback to the byte copy loop.

Despite what ca96ab859ab4 ("lib/sort: Add 64 bit swap function") claims,
very few users of sort() sort pointers (or pointer-sized objects); most
sort structures containing at least two words.  (E.g.
drivers/acpi/fan.c:acpi_fan_get_fps() sorts an array of 40-byte struct
acpi_fan_fps.)

The functions also got renamed to reflect the fact that they support
multiple words.  In the great tradition of bikeshedding, the names were
by far the most contentious issue during review of this patch series.

x86-64 code size 872 -> 886 bytes (+14)

With feedback from Andy Shevchenko, Rasmus Villemoes and Geert
Uytterhoeven.

Link: http://lkml.kernel.org/r/f24f932df3a7fa1973c1084154f1cea596bcf341.1552704200.git.lkml@sdf.org
Signed-off-by: George Spelvin <lkml@sdf.org>
Acked-by: Andrey Abramov <st5pub@yandex.ru>
Acked-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Daniel Wagner <daniel.wagner@siemens.com>
Cc: Don Mullis <don.mullis@gmail.com>
Cc: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

lib/plist: rename DEBUG_PI_LIST to DEBUG_PLIST

This is a lot more appropriate than PI_LIST, which in the kernel one
would assume that it has to do with priority-inheritance; which is not
-- furthermore futexes make use of plists so this can be even more
confusing, albeit the debug nature of the config option.

Link: http://lkml.kernel.org/r/20190317185434.1626-1-dave@stgolabs.net
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

lib/bitmap.c: guard exotic bitmap functions by CONFIG_NUMA

The bitmap_remap, _bitremap, _onto and _fold functions are only used,
via their node_ wrappers, in mm/mempolicy.c, which is only built for
CONFIG_NUMA.  The helper bitmap_ord_to_pos used by these functions is
global, but its only external caller is node_random() in lib/nodemask.c,
which is also guarded by CONFIG_NUMA.

For !CONFIG_NUMA:

add/remove: 0/6 grow/shrink: 0/0 up/down: 0/-621 (-621)
Function                                     old     new   delta
bitmap_pos_to_ord                             20       -     -20
bitmap_ord_to_pos                             70       -     -70
bitmap_bitremap                               81       -     -81
bitmap_fold                                  113       -    -113
bitmap_onto                                  123       -    -123
bitmap_remap                                 214       -    -214
Total: Before=4776, After=4155, chg -13.00%

Link: http://lkml.kernel.org/r/20190329205353.6010-2-linux@rasmusvillemoes.dk
Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Cc: Andy Shevchenko <andy.shevchenko@gmail.com>
Cc: Yury Norov <yury.norov@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

lib/bitmap.c: remove unused EXPORT_SYMBOLs

AFAICT, there have never been any callers of these functions outside
mm/mempolicy.c (via their nodemask.h wrappers).  In particular, no
modular code has ever used them, and given their somewhat exotic
semantics, I highly doubt they will ever find such a use.  In any case,
no need to export them currently.

Link: http://lkml.kernel.org/r/20190329205353.6010-1-linux@rasmusvillemoes.dk
Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Cc: Andy Shevchenko <andy.shevchenko@gmail.com>
Cc: Yury Norov <yury.norov@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

kernel/user.c: clean up some leftover code

The out_unlock label is misleading; no unlocking happens after it, so
just return NULL directly.

Also, nothing between the kmem_cache_zalloc() that creates new and the
two key_put() can initialize new->uid_keyring or new->session_keyring,
so those calls are no-ops.

Link: http://lkml.kernel.org/r/20190424200404.9114-1-linux@rasmusvillemoes.dk
Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: "Peter Zijlstra (Intel)" <peterz@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

kernel/latencytop.c: rename clear_all_latency_tracing to clear_tsk_latency_tracing

The name clear_all_latency_tracing is misleading, in fact which only
clear per task's latency_record[], and we do have another function named
clear_global_latency_tracing which clear the global latency_record[]
buffer.

Link: http://lkml.kernel.org/r/20190226114602.16902-1-linf@wangsu.com
Signed-off-by: Lin Feng <linf@wangsu.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Fabian Frederick <fabf@skynet.be>
Cc: Arjan van de Ven <arjan@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

kernel/latencytop.c: remove unnecessary checks for latencytop_enabled

1. In latencytop source codes, we only have such calling chain:

account_scheduler_latency(struct task_struct *task, int usecs, int inter)
{
        if (unlikely(latencytop_enabled)) /* the outtermost check */
                __account_scheduler_latency(task, usecs, inter);
}
__account_scheduler_latency
    account_global_scheduler_latency
        if (!latencytop_enabled)

So, the inner check for latencytop_enabled is not necessary at all.

2. In clear_all_latency_tracing and now is called
   clear_tsk_latency_tracing the check for latencytop_enabled is redundant
   and buggy to some extent.

   We have no reason to refuse clearing the /proc/$pid/latency if
   latencytop_enabled is set to 0, considering that if we use latencytop
   manually by echo 0 > /proc/sys/kernel/latencytop, then we want to clear
   /proc/$pid/latency and failed.

   Also we don't have such check in brother function
   clear_global_latency_tracing.

Notes: These changes are only visible to users who set
   CONFIG_LATENCYTOP and won't change user tool latencytop's behavior.

Link: http://lkml.kernel.org/r/20190226114602.16902-2-linf@wangsu.com
Signed-off-by: Lin Feng <linf@wangsu.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Fabian Frederick <fabf@skynet.be>
Cc: Arjan van de Ven <arjan@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

kernel/notifier.c: double register detection

By design notifiers can be registerd once only, 2nd register attempt
called by mistake silently corrupts notifiers list.

A few years ago I investigated described problem, the host was power
cycled because of notifier list corruption.  I've prepared this patch
and applied it to the OpenVZ kernel and sent this patch but nobody
commented on it.  Later it helped us to detect a similar problem in the
OpenVz kernel.

Mistakes with notifier registration can happen for example during
subsystem initialization from different namespaces, or because of a lost
unregister in the roll-back path on initialization failures.

The proposed check cannot prevent the described problem, however it
allows us to detect its reason quickly without coredump analysis.

Link: http://lkml.kernel.org/r/04127e71-4782-9bbb-fe5a-7c01e93a99b0@virtuozzo.com
Signed-off-by: Vasily Averin <vvs@virtuozzo.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

compiler: allow all arches to enable CONFIG_OPTIMIZE_INLINING

Commit 60a3cdd06394 ("x86: add optimized inlining") introduced
CONFIG_OPTIMIZE_INLINING, but it has been available only for x86.

The idea is obviously arch-agnostic.  This commit moves the config entry
from arch/x86/Kconfig.debug to lib/Kconfig.debug so that all
architectures can benefit from it.

This can make a huge difference in kernel image size especially when
CONFIG_OPTIMIZE_FOR_SIZE is enabled.

For example, I got 3.5% smaller arm64 kernel for v5.1-rc1.

  dec       file
  18983424  arch/arm64/boot/Image.before
  18321920  arch/arm64/boot/Image.after

This also slightly improves the "Kernel hacking" Kconfig menu as
e61aca5158a8 ("Merge branch 'kconfig-diet' from Dave Hansen') suggested;
this config option would be a good fit in the "compiler option" menu.

Link: http://lkml.kernel.org/r/20190423034959.13525-12-yamada.masahiro@socionext.com
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Acked-by: Borislav Petkov <bp@suse.de>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Boris Brezillon <bbrezillon@kernel.org>
Cc: Brian Norris <computersforpeace@gmail.com>
Cc: Christophe Leroy <christophe.leroy@c-s.fr>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Marek Vasut <marek.vasut@gmail.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mathieu Malaterre <malat@debian.org>
Cc: Miquel Raynal <miquel.raynal@bootlin.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Richard Weinberger <richard@nod.at>
Cc: Russell King <rmk+kernel@arm.linux.org.uk>
Cc: Stefan Agner <stefan@agner.ch>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

powerpc/mm/radix: mark as __tlbie_pid() and friends as__always_inline

This prepares to move CONFIG_OPTIMIZE_INLINING from x86 to a common
place.  We need to eliminate potential issues beforehand.

If it is enabled for powerpc, the following errors are reported:

  arch/powerpc/mm/tlb-radix.c: In function '__tlbie_lpid':
  arch/powerpc/mm/tlb-radix.c:148:2: warning: asm operand 3 probably doesn't match constraints
    asm volatile(PPC_TLBIE_5(%0, %4, %3, %2, %1)
    ^~~
  arch/powerpc/mm/tlb-radix.c:148:2: error: impossible constraint in 'asm'
  arch/powerpc/mm/tlb-radix.c: In function '__tlbie_pid':
  arch/powerpc/mm/tlb-radix.c:118:2: warning: asm operand 3 probably doesn't match constraints
    asm volatile(PPC_TLBIE_5(%0, %4, %3, %2, %1)
    ^~~
  arch/powerpc/mm/tlb-radix.c: In function '__tlbiel_pid':
  arch/powerpc/mm/tlb-radix.c:104:2: warning: asm operand 3 probably doesn't match constraints
    asm volatile(PPC_TLBIEL(%0, %4, %3, %2, %1)
    ^~~

Link: http://lkml.kernel.org/r/20190423034959.13525-11-yamada.masahiro@socionext.com
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Boris Brezillon <bbrezillon@kernel.org>
Cc: Borislav Petkov <bp@suse.de>
Cc: Brian Norris <computersforpeace@gmail.com>
Cc: Christophe Leroy <christophe.leroy@c-s.fr>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Marek Vasut <marek.vasut@gmail.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mathieu Malaterre <malat@debian.org>
Cc: Miquel Raynal <miquel.raynal@bootlin.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Richard Weinberger <richard@nod.at>
Cc: Russell King <rmk+kernel@arm.linux.org.uk>
Cc: Stefan Agner <stefan@agner.ch>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

powerpc/mm/radix: mark __radix__flush_tlb_range_psize() as __always_inline

This prepares to move CONFIG_OPTIMIZE_INLINING from x86 to a common
place.  We need to eliminate potential issues beforehand.

If it is enabled for powerpc, the following error is reported:

  arch/powerpc/mm/tlb-radix.c: In function '__radix__flush_tlb_range_psize':
  arch/powerpc/mm/tlb-radix.c:104:2: error: asm operand 3 probably doesn't match constraints [-Werror]
    asm volatile(PPC_TLBIEL(%0, %4, %3, %2, %1)
    ^~~
  arch/powerpc/mm/tlb-radix.c:104:2: error: impossible constraint in 'asm'

Link: http://lkml.kernel.org/r/20190423034959.13525-10-yamada.masahiro@socionext.com
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Boris Brezillon <bbrezillon@kernel.org>
Cc: Borislav Petkov <bp@suse.de>
Cc: Brian Norris <computersforpeace@gmail.com>
Cc: Christophe Leroy <christophe.leroy@c-s.fr>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Marek Vasut <marek.vasut@gmail.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mathieu Malaterre <malat@debian.org>
Cc: Miquel Raynal <miquel.raynal@bootlin.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Richard Weinberger <richard@nod.at>
Cc: Russell King <rmk+kernel@arm.linux.org.uk>
Cc: Stefan Agner <stefan@agner.ch>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

powerpc/prom_init: mark prom_getprop() and prom_getproplen() as __init

This prepares to move CONFIG_OPTIMIZE_INLINING from x86 to a common
place.  We need to eliminate potential issues beforehand.

If it is enabled for powerpc, the following modpost warnings are
reported:

  WARNING: vmlinux.o(.text.unlikely+0x20): Section mismatch in reference from the function .prom_getprop() to the function .init.text:.call_prom()
  The function .prom_getprop() references the function __init .call_prom().
  This is often because .prom_getprop lacks a __init annotation or the annotation of .call_prom is wrong.

  WARNING: vmlinux.o(.text.unlikely+0x3c): Section mismatch in reference from the function .prom_getproplen() to the function .init.text:.call_prom()
  The function .prom_getproplen() references the function __init .call_prom().
  This is often because .prom_getproplen lacks a __init annotation or the annotation of .call_prom is wrong.

Link: http://lkml.kernel.org/r/20190423034959.13525-9-yamada.masahiro@socionext.com
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Boris Brezillon <bbrezillon@kernel.org>
Cc: Borislav Petkov <bp@suse.de>
Cc: Brian Norris <computersforpeace@gmail.com>
Cc: Christophe Leroy <christophe.leroy@c-s.fr>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Marek Vasut <marek.vasut@gmail.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mathieu Malaterre <malat@debian.org>
Cc: Miquel Raynal <miquel.raynal@bootlin.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Richard Weinberger <richard@nod.at>
Cc: Russell King <rmk+kernel@arm.linux.org.uk>
Cc: Stefan Agner <stefan@agner.ch>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ARM: mark setup_machine_tags() stub as __init __noreturn

This prepares to move CONFIG_OPTIMIZE_INLINING from x86 to a common
place.  We need to eliminate potential issues beforehand.

If it is enabled for arm, Clang build results in the following modpost
warning:

  WARNING: vmlinux.o(.text+0x1124): Section mismatch in reference from the function setup_machine_tags() to the function .init.text:early_print()
  The function setup_machine_tags() references the function __init early_print().
  This is often because setup_machine_tags lacks a __init annotation or the annotation of early_print is wrong.

Link: http://lkml.kernel.org/r/20190423034959.13525-8-yamada.masahiro@socionext.com
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Boris Brezillon <bbrezillon@kernel.org>
Cc: Borislav Petkov <bp@suse.de>
Cc: Brian Norris <computersforpeace@gmail.com>
Cc: Christophe Leroy <christophe.leroy@c-s.fr>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Marek Vasut <marek.vasut@gmail.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mathieu Malaterre <malat@debian.org>
Cc: Miquel Raynal <miquel.raynal@bootlin.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Richard Weinberger <richard@nod.at>
Cc: Russell King <rmk+kernel@arm.linux.org.uk>
Cc: Stefan Agner <stefan@agner.ch>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

MIPS: mark __fls() and __ffs() as __always_inline

This prepares to move CONFIG_OPTIMIZE_INLINING from x86 to a common
place.  We need to eliminate potential issues beforehand.

If it is enabled for mips, the following errors are reported:

  arch/mips/mm/sc-mips.o: In function `mips_sc_prefetch_enable.part.2':
  sc-mips.c:(.text+0x98): undefined reference to `mips_gcr_base'
  sc-mips.c:(.text+0x9c): undefined reference to `mips_gcr_base'
  sc-mips.c:(.text+0xbc): undefined reference to `mips_gcr_base'
  sc-mips.c:(.text+0xc8): undefined reference to `mips_gcr_base'
  sc-mips.c:(.text+0xdc): undefined reference to `mips_gcr_base'
  arch/mips/mm/sc-mips.o:sc-mips.c:(.text.unlikely+0x44): more undefined references to `mips_gcr_base'

Link: http://lkml.kernel.org/r/20190423034959.13525-7-yamada.masahiro@socionext.com
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Boris Brezillon <bbrezillon@kernel.org>
Cc: Borislav Petkov <bp@suse.de>
Cc: Brian Norris <computersforpeace@gmail.com>
Cc: Christophe Leroy <christophe.leroy@c-s.fr>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Marek Vasut <marek.vasut@gmail.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mathieu Malaterre <malat@debian.org>
Cc: Miquel Raynal <miquel.raynal@bootlin.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Richard Weinberger <richard@nod.at>
Cc: Russell King <rmk+kernel@arm.linux.org.uk>
Cc: Stefan Agner <stefan@agner.ch>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mtd: rawnand: vf610_nfc: add initializer to avoid -Wmaybe-uninitialized

This prepares to move CONFIG_OPTIMIZE_INLINING from x86 to a common
place.  We need to eliminate potential issues beforehand.

Kbuild test robot has never reported -Wmaybe-uninitialized warning for
this probably because vf610_nfc_run() is inlined by the x86 compiler's
inlining heuristic.

If CONFIG_OPTIMIZE_INLINING is enabled for a different architecture and
vf610_nfc_run() is not inlined, the following warning is reported:

  drivers/mtd/nand/raw/vf610_nfc.c: In function `vf610_nfc_cmd':
  drivers/mtd/nand/raw/vf610_nfc.c:455:3: warning: `offset' may be used uninitialized in this function [-Wmaybe-uninitialized]
     vf610_nfc_rd_from_sram(instr->ctx.data.buf.in + offset,
     ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
              nfc->regs + NFC_MAIN_AREA(0) + offset,
              ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
              trfr_sz, !nfc->data_access);
              ~~~~~~~~~~~~~~~~~~~~~~~~~~~

Link: http://lkml.kernel.org/r/20190423034959.13525-6-yamada.masahiro@socionext.com
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Boris Brezillon <bbrezillon@kernel.org>
Cc: Borislav Petkov <bp@suse.de>
Cc: Brian Norris <computersforpeace@gmail.com>
Cc: Christophe Leroy <christophe.leroy@c-s.fr>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Marek Vasut <marek.vasut@gmail.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mathieu Malaterre <malat@debian.org>
Cc: Miquel Raynal <miquel.raynal@bootlin.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Richard Weinberger <richard@nod.at>
Cc: Russell King <rmk+kernel@arm.linux.org.uk>
Cc: Stefan Agner <stefan@agner.ch>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

s390/cpacf: mark scpacf_query() as __always_inline

This prepares to move CONFIG_OPTIMIZE_INLINING from x86 to a common
place.  We need to eliminate potential issues beforehand.

If it is enabled for s390, the following error is reported:

  In file included from arch/s390/crypto/des_s390.c:19:
  arch/s390/include/asm/cpacf.h: In function 'cpacf_query':
  arch/s390/include/asm/cpacf.h:170:2: warning: asm operand 3 probably doesn't match constraints
    asm volatile(
    ^~~
  arch/s390/include/asm/cpacf.h:170:2: error: impossible constraint in 'asm'

Link: http://lkml.kernel.org/r/20190423034959.13525-5-yamada.masahiro@socionext.com
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Boris Brezillon <bbrezillon@kernel.org>
Cc: Borislav Petkov <bp@suse.de>
Cc: Brian Norris <computersforpeace@gmail.com>
Cc: Christophe Leroy <christophe.leroy@c-s.fr>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Marek Vasut <marek.vasut@gmail.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mathieu Malaterre <malat@debian.org>
Cc: Miquel Raynal <miquel.raynal@bootlin.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Richard Weinberger <richard@nod.at>
Cc: Russell King <rmk+kernel@arm.linux.org.uk>
Cc: Stefan Agner <stefan@agner.ch>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

MIPS: mark mult_sh_align_mod() as __always_inline

This prepares to move CONFIG_OPTIMIZE_INLINING from x86 to a common
place.  We need to eliminate potential issues beforehand.

If it is enabled for mips, the following error is reported:

  arch/mips/kernel/cpu-bugs64.c: In function 'mult_sh_align_mod.constprop':
  arch/mips/kernel/cpu-bugs64.c:33:2: error: asm operand 1 probably doesn't match constraints [-Werror]
    asm volatile(
    ^~~
  arch/mips/kernel/cpu-bugs64.c:33:2: error: asm operand 1 probably doesn't match constraints [-Werror]
    asm volatile(
    ^~~
  arch/mips/kernel/cpu-bugs64.c:33:2: error: impossible constraint in 'asm'
    asm volatile(
    ^~~
  arch/mips/kernel/cpu-bugs64.c:33:2: error: impossible constraint in 'asm'
    asm volatile(
    ^~~

Link: http://lkml.kernel.org/r/20190423034959.13525-4-yamada.masahiro@socionext.com
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Boris Brezillon <bbrezillon@kernel.org>
Cc: Borislav Petkov <bp@suse.de>
Cc: Brian Norris <computersforpeace@gmail.com>
Cc: Christophe Leroy <christophe.leroy@c-s.fr>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Marek Vasut <marek.vasut@gmail.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mathieu Malaterre <malat@debian.org>
Cc: Miquel Raynal <miquel.raynal@bootlin.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Richard Weinberger <richard@nod.at>
Cc: Russell King <rmk+kernel@arm.linux.org.uk>
Cc: Stefan Agner <stefan@agner.ch>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

arm64: mark (__)cpus_have_const_cap as __always_inline

This prepares to move CONFIG_OPTIMIZE_INLINING from x86 to a common
place.  We need to eliminate potential issues beforehand.

If it is enabled for arm64, the following errors are reported:

  In file included from include/linux/compiler_types.h:68,
                   from <command-line>:
  arch/arm64/include/asm/jump_label.h: In function 'cpus_have_const_cap':
  include/linux/compiler-gcc.h:120:38: warning: asm operand 0 probably doesn't match constraints
   #define asm_volatile_goto(x...) do { asm goto(x); asm (""); } while (0)
                                        ^~~
  arch/arm64/include/asm/jump_label.h:32:2: note: in expansion of macro 'asm_volatile_goto'
    asm_volatile_goto(
    ^~~~~~~~~~~~~~~~~
  include/linux/compiler-gcc.h:120:38: error: impossible constraint in 'asm'
   #define asm_volatile_goto(x...) do { asm goto(x); asm (""); } while (0)
                                        ^~~
  arch/arm64/include/asm/jump_label.h:32:2: note: in expansion of macro 'asm_volatile_goto'
    asm_volatile_goto(
    ^~~~~~~~~~~~~~~~~

Link: http://lkml.kernel.org/r/20190423034959.13525-3-yamada.masahiro@socionext.com
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Tested-by: Mark Rutland <mark.rutland@arm.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Boris Brezillon <bbrezillon@kernel.org>
Cc: Borislav Petkov <bp@suse.de>
Cc: Brian Norris <computersforpeace@gmail.com>
Cc: Christophe Leroy <christophe.leroy@c-s.fr>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Marek Vasut <marek.vasut@gmail.com>
Cc: Mathieu Malaterre <malat@debian.org>
Cc: Miquel Raynal <miquel.raynal@bootlin.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Richard Weinberger <richard@nod.at>
Cc: Russell King <rmk+kernel@arm.linux.org.uk>
Cc: Stefan Agner <stefan@agner.ch>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ARM: prevent tracing IPI_CPU_BACKTRACE

Patch series "compiler: allow all arches to enable
CONFIG_OPTIMIZE_INLINING", v3.

This patch (of 11):

When function tracing for IPIs is enabled, we get a warning for an
overflow of the ipi_types array with the IPI_CPU_BACKTRACE type as
triggered by raise_nmi():

  arch/arm/kernel/smp.c: In function 'raise_nmi':
  arch/arm/kernel/smp.c:489:2: error: array subscript is above array bounds [-Werror=array-bounds]
    trace_ipi_raise(target, ipi_types[ipinr]);

This is a correct warning as we actually overflow the array here.

This patch raise_nmi() to call __smp_cross_call() instead of
smp_cross_call(), to avoid calling into ftrace.  For clarification, I'm
also adding a two new code comments describing how this one is special.

The warning appears to have shown up after commit e7273ff49acf ("ARM:
8488/1: Make IPI_CPU_BACKTRACE a "non-secure" SGI"), which changed the
number assignment from '15' to '8', but as far as I can tell has existed
since the IPI tracepoints were first introduced.  If we decide to
backport this patch to stable kernels, we probably need to backport
e7273ff49acf as well.

[yamada.masahiro@socionext.com: rebase on v5.1-rc1]
Link: http://lkml.kernel.org/r/20190423034959.13525-2-yamada.masahiro@socionext.com
Fixes: e7273ff49acf ("ARM: 8488/1: Make IPI_CPU_BACKTRACE a "non-secure" SGI")
Fixes: 365ec7b17327 ("ARM: add IPI tracepoints") # v3.17
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Christophe Leroy <christophe.leroy@c-s.fr>
Cc: Mathieu Malaterre <malat@debian.org>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Stefan Agner <stefan@agner.ch>
Cc: Boris Brezillon <bbrezillon@kernel.org>
Cc: Miquel Raynal <miquel.raynal@bootlin.com>
Cc: Richard Weinberger <richard@nod.at>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Brian Norris <computersforpeace@gmail.com>
Cc: Marek Vasut <marek.vasut@gmail.com>
Cc: Russell King <rmk+kernel@arm.linux.org.uk>
Cc: Borislav Petkov <bp@suse.de>
Cc: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

treewide: remove SPDX "WITH Linux-syscall-note" from kernel-space headers

The "WITH Linux-syscall-note" should be added to headers exported to the
user-space.

Some kernel-space headers have "WITH Linux-syscall-note", which seems a
mistake.

[1] arch/x86/include/asm/hyperv-tlfs.h

Commit 5a4858032217 ("x86/hyper-v: move hyperv.h out of uapi") moved
this file out of uapi, but missed to update the SPDX License tag.

[2] include/asm-generic/shmparam.h

Commit 76ce2a80a28e ("Rename include/{uapi => }/asm-generic/shmparam.h
really") moved this file out of uapi, but missed to update the SPDX
License tag.

[3] include/linux/qcom-geni-se.h

Commit eddac5af0654 ("soc: qcom: Add GENI based QUP Wrapper driver")
added this file, but I do not see a good reason why its license tag must
include "WITH Linux-syscall-note".

Link: http://lkml.kernel.org/r/1554196104-3522-1-git-send-email-yamada.masahiro@socionext.com
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fs/select: avoid clang stack usage warning

The select() implementation is carefully tuned to put a sensible amount
of data on the stack for holding a copy of the user space fd_set, but
not too large to risk overflowing the kernel stack.

When building a 32-bit kernel with clang, we need a little more space
than with gcc, which often triggers a warning:

  fs/select.c:619:5: error: stack frame size of 1048 bytes in function 'core_sys_select' [-Werror,-Wframe-larger-than=]
  int core_sys_select(int n, fd_set __user *inp, fd_set __user *outp,

I experimentally found that for 32-bit ARM, reducing the maximum stack
usage by 64 bytes keeps us reliably under the warning limit again.

Link: http://lkml.kernel.org/r/20190307090146.1874906-1-arnd@arndb.de
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Andi Kleen <ak@linux.intel.com>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Eric Dumazet <edumazet@google.com>
Cc: "Darrick J. Wong" <darrick.wong@oracle.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/mincore.c: make mincore() more conservative

The semantics of what mincore() considers to be resident is not
completely clear, but Linux has always (since 2.3.52, which is when
mincore() was initially done) treated it as "page is available in page
cache".

That's potentially a problem, as that [in]directly exposes
meta-information about pagecache / memory mapping state even about
memory not strictly belonging to the process executing the syscall,
opening possibilities for sidechannel attacks.

Change the semantics of mincore() so that it only reveals pagecache
information for non-anonymous mappings that belog to files that the
calling process could (if it tried to) successfully open for writing;
otherwise we'd be including shared non-exclusive mappings, which

 - is the sidechannel

 - is not the usecase for mincore(), as that's primarily used for data,
   not (shared) text

[jkosina@suse.cz: v2]
Link: http://lkml.kernel.org/r/20190312141708.6652-2-vbabka@suse.cz
[mhocko@suse.com: restructure can_do_mincore() conditions]
Link: http://lkml.kernel.org/r/nycvar.YFH.7.76.1903062342020.19912@cbobk.fhfr.pm
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Josh Snyder <joshs@netflix.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Originally-by: Linus Torvalds <torvalds@linux-foundation.org>
Originally-by: Dominique Martinet <asmadeus@codewreck.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Kevin Easton <kevin@guarana.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Cyril Hrubis <chrubis@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Kirill A. Shutemov <kirill@shutemov.name>
Cc: Daniel Gruss <daniel@gruss.cc>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: maintain randomization of page free lists

When freeing a page with an order >= shuffle_page_order randomly select
the front or back of the list for insertion.

While the mm tries to defragment physical pages into huge pages this can
tend to make the page allocator more predictable over time.  Inject the
front-back randomness to preserve the initial randomness established by
shuffle_free_memory() when the kernel was booted.

The overhead of this manipulation is constrained by only being applied
for MAX_ORDER sized pages by default.

[akpm@linux-foundation.org: coding-style fixes]
Link: http://lkml.kernel.org/r/154899812788.3165233.9066631950746578517.stgit@dwillia2-desk3.amr.corp.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Keith Busch <keith.busch@intel.com>
Cc: Robert Elliott <elliott@hpe.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: move buddy list manipulations into helpers

In preparation for runtime randomization of the zone lists, take all
(well, most of) the list_*() functions in the buddy allocator and put
them in helper functions.  Provide a common control point for injecting
additional behavior when freeing pages.

[dan.j.williams@intel.com: fix buddy list helpers]
Link: http://lkml.kernel.org/r/155033679702.1773410.13041474192173212653.stgit@dwillia2-desk3.amr.corp.intel.com
[vbabka@suse.cz: remove del_page_from_free_area() migratetype parameter]
Link: http://lkml.kernel.org/r/4672701b-6775-6efd-0797-b6242591419e@suse.cz
Link: http://lkml.kernel.org/r/154899812264.3165233.5219320056406926223.stgit@dwillia2-desk3.amr.corp.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Tested-by: Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Keith Busch <keith.busch@intel.com>
Cc: Robert Elliott <elliott@hpe.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: shuffle initial free memory to improve memory-side-cache utilization

Patch series "mm: Randomize free memory", v10.

This patch (of 3):

Randomization of the page allocator improves the average utilization of
a direct-mapped memory-side-cache.  Memory side caching is a platform
capability that Linux has been previously exposed to in HPC
(high-performance computing) environments on specialty platforms.  In
that instance it was a smaller pool of high-bandwidth-memory relative to
higher-capacity / lower-bandwidth DRAM.  Now, this capability is going
to be found on general purpose server platforms where DRAM is a cache in
front of higher latency persistent memory [1].

Robert offered an explanation of the state of the art of Linux
interactions with memory-side-caches [2], and I copy it here:

    It's been a problem in the HPC space:
    http://www.nersc.gov/research-and-development/knl-cache-mode-performance-coe/

    A kernel module called zonesort is available to try to help:
    https://software.intel.com/en-us/articles/xeon-phi-software

    and this abandoned patch series proposed that for the kernel:
    https://lkml.kernel.org/r/20170823100205.17311-1-lukasz.daniluk@intel.com

    Dan's patch series doesn't attempt to ensure buffers won't conflict, but
    also reduces the chance that the buffers will. This will make performance
    more consistent, albeit slower than "optimal" (which is near impossible
    to attain in a general-purpose kernel).  That's better than forcing
    users to deploy remedies like:
        "To eliminate this gradual degradation, we have added a Stream
         measurement to the Node Health Check that follows each job;
         nodes are rebooted whenever their measured memory bandwidth
         falls below 300 GB/s."

A replacement for zonesort was merged upstream in commit cc9aec03e58f
("x86/numa_emulation: Introduce uniform split capability").  With this
numa_emulation capability, memory can be split into cache sized
("near-memory" sized) numa nodes.  A bind operation to such a node, and
disabling workloads on other nodes, enables full cache performance.
However, once the workload exceeds the cache size then cache conflicts
are unavoidable.  While HPC environments might be able to tolerate
time-scheduling of cache sized workloads, for general purpose server
platforms, the oversubscribed cache case will be the common case.

The worst case scenario is that a server system owner benchmarks a
workload at boot with an un-contended cache only to see that performance
degrade over time, even below the average cache performance due to
excessive conflicts.  Randomization clips the peaks and fills in the
valleys of cache utilization to yield steady average performance.

Here are some performance impact details of the patches:

1/ An Intel internal synthetic memory bandwidth measurement tool, saw a
   3X speedup in a contrived case that tries to force cache conflicts.
   The contrived cased used the numa_emulation capability to force an
   instance of the benchmark to be run in two of the near-memory sized
   numa nodes.  If both instances were placed on the same emulated they
   would fit and cause zero conflicts.  While on separate emulated nodes
   without randomization they underutilized the cache and conflicted
   unnecessarily due to the in-order allocation per node.

2/ A well known Java server application benchmark was run with a heap
   size that exceeded cache size by 3X.  The cache conflict rate was 8%
   for the first run and degraded to 21% after page allocator aging.  With
   randomization enabled the rate levelled out at 11%.

3/ A MongoDB workload did not observe measurable difference in
   cache-conflict rates, but the overall throughput dropped by 7% with
   randomization in one case.

4/ Mel Gorman ran his suite of performance workloads with randomization
   enabled on platforms without a memory-side-cache and saw a mix of some
   improvements and some losses [3].

While there is potentially significant improvement for applications that
depend on low latency access across a wide working-set, the performance
may be negligible to negative for other workloads.  For this reason the
shuffle capability defaults to off unless a direct-mapped
memory-side-cache is detected.  Even then, the page_alloc.shuffle=0
parameter can be specified to disable the randomization on those systems.

Outside of memory-side-cache utilization concerns there is potentially
security benefit from randomization.  Some data exfiltration and
return-oriented-programming attacks rely on the ability to infer the
location of sensitive data objects.  The kernel page allocator, especially
early in system boot, has predictable first-in-first out behavior for
physical pages.  Pages are freed in physical address order when first
onlined.

Quoting Kees:
    "While we already have a base-address randomization
     (CONFIG_RANDOMIZE_MEMORY), attacks against the same hardware and
     memory layouts would certainly be using the predictability of
     allocation ordering (i.e. for attacks where the base address isn't
     important: only the relative positions between allocated memory).
     This is common in lots of heap-style attacks. They try to gain
     control over ordering by spraying allocations, etc.

     I'd really like to see this because it gives us something similar
     to CONFIG_SLAB_FREELIST_RANDOM but for the page allocator."

While SLAB_FREELIST_RANDOM reduces the predictability of some local slab
caches it leaves vast bulk of memory to be predictably in order allocated.
However, it should be noted, the concrete security benefits are hard to
quantify, and no known CVE is mitigated by this randomization.

Introduce shuffle_free_memory(), and its helper shuffle_zone(), to perform
a Fisher-Yates shuffle of the page allocator 'free_area' lists when they
are initially populated with free memory at boot and at hotplug time.  Do
this based on either the presence of a page_alloc.shuffle=Y command line
parameter, or autodetection of a memory-side-cache (to be added in a
follow-on patch).

The shuffling is done in terms of CONFIG_SHUFFLE_PAGE_ORDER sized free
pages where the default CONFIG_SHUFFLE_PAGE_ORDER is MAX_ORDER-1 i.e.  10,
4MB this trades off randomization granularity for time spent shuffling.
MAX_ORDER-1 was chosen to be minimally invasive to the page allocator
while still showing memory-side cache behavior improvements, and the
expectation that the security implications of finer granularity
randomization is mitigated by CONFIG_SLAB_FREELIST_RANDOM.  The
performance impact of the shuffling appears to be in the noise compared to
other memory initialization work.

This initial randomization can be undone over time so a follow-on patch is
introduced to inject entropy on page free decisions.  It is reasonable to
ask if the page free entropy is sufficient, but it is not enough due to
the in-order initial freeing of pages.  At the start of that process
putting page1 in front or behind page0 still keeps them close together,
page2 is still near page1 and has a high chance of being adjacent.  As
more pages are added ordering diversity improves, but there is still high
page locality for the low address pages and this leads to no significant
impact to the cache conflict rate.

[1]: https://itpeernetwork.intel.com/intel-optane-dc-persistent-memory-operating-modes/
[2]: https://lkml.kernel.org/r/AT5PR8401MB1169D656C8B5E121752FC0F8AB120@AT5PR8401MB1169.NAMPRD84.PROD.OUTLOOK.COM
[3]: https://lkml.org/lkml/2018/10/12/309

[dan.j.williams@intel.com: fix shuffle enable]
Link: http://lkml.kernel.org/r/154943713038.3858443.4125180191382062871.stgit@dwillia2-desk3.amr.corp.intel.com
[cai@lca.pw: fix SHUFFLE_PAGE_ALLOCATOR help texts]
Link: http://lkml.kernel.org/r/20190425201300.75650-1-cai@lca.pw
Link: http://lkml.kernel.org/r/154899811738.3165233.12325692939590944259.stgit@dwillia2-desk3.amr.corp.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Qian Cai <cai@lca.pw>
Reviewed-by: Kees Cook <keescook@chromium.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Keith Busch <keith.busch@intel.com>
Cc: Robert Elliott <elliott@hpe.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/vmalloc.c: convert vmap_lazy_nr to atomic_long_t

vmap_lazy_nr variable has atomic_t type that is 4 bytes integer value on
both 32 and 64 bit systems.  lazy_max_pages() deals with "unsigned long"
that is 8 bytes on 64 bit system, thus vmap_lazy_nr should be 8 bytes on
64 bit as well.

Link: http://lkml.kernel.org/r/20190131162452.25879-1-urezki@gmail.com
Signed-off-by: Uladzislau Rezki (Sony) <urezki@gmail.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Thomas Garnier <thgarnie@google.com>
Cc: Oleksiy Avramchenko <oleksiy.avramchenko@sonymobile.com>
Cc: Joel Fernandes <joelaf@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/vmalloc.c: add priority threshold to __purge_vmap_area_lazy()

Commit 763b218ddfaf ("mm: add preempt points into __purge_vmap_area_lazy()")
introduced some preempt points, one of those is making an allocation
more prioritized over lazy free of vmap areas.

Prioritizing an allocation over freeing does not work well all the time,
i.e.  it should be rather a compromise.

1) Number of lazy pages directly influences the busy list length thus
   on operations like: allocation, lookup, unmap, remove, etc.

2) Under heavy stress of vmalloc subsystem I run into a situation when
   memory usage gets increased hitting out_of_memory -> panic state due to
   completely blocking of logic that frees vmap areas in the
   __purge_vmap_area_lazy() function.

Establish a threshold passing which the freeing is prioritized back over
allocation creating a balance between each other.

Using vmalloc test driver in "stress mode", i.e.  When all available
test cases are run simultaneously on all online CPUs applying a
pressure on the vmalloc subsystem, my HiKey 960 board runs out of
memory due to the fact that __purge_vmap_area_lazy() logic simply is
not able to free pages in time.

How I run it:

1) You should build your kernel with CONFIG_TEST_VMALLOC=m
2) ./tools/testing/selftests/vm/test_vmalloc.sh stress

During this test "vmap_lazy_nr" pages will go far beyond acceptable
lazy_max_pages() threshold, that will lead to enormous busy list size
and other problems including allocation time and so on.

Link: http://lkml.kernel.org/r/20190124115648.9433-3-urezki@gmail.com
Signed-off-by: Uladzislau Rezki (Sony) <urezki@gmail.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Thomas Garnier <thgarnie@google.com>
Cc: Oleksiy Avramchenko <oleksiy.avramchenko@sonymobile.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Joel Fernandes <joelaf@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Tejun Heo <tj@kernel.org>
Cc: Joel Fernandes <joel@joelfernandes.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

kernel/sched/psi.c: expose pressure metrics on root cgroup

Pressure metrics are already recorded and exposed in procfs for the
entire system, but any tool which monitors cgroup pressure has to
special case the root cgroup to read from procfs.  This patch exposes
the already recorded pressure metrics on the root cgroup.

Link: http://lkml.kernel.org/r/20190510174938.3361741-1-dschatzberg@fb.com
Signed-off-by: Dan Schatzberg <dschatzberg@fb.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Tejun Heo <tj@kernel.org>
Cc: Li Zefan <lizefan@huawei.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

psi: introduce psi monitor

Psi monitor aims to provide a low-latency short-term pressure detection
mechanism configurable by users.  It allows users to monitor psi metrics
growth and trigger events whenever a metric raises above user-defined
threshold within user-defined time window.

Time window and threshold are both expressed in usecs.  Multiple psi
resources with different thresholds and window sizes can be monitored
concurrently.

Psi monitors activate when system enters stall state for the monitored
psi metric and deactivate upon exit from the stall state.  While system
is in the stall state psi signal growth is monitored at a rate of 10
times per tracking window.  Min window size is 500ms, therefore the min
monitoring interval is 50ms.  Max window size is 10s with monitoring
interval of 1s.

When activated psi monitor stays active for at least the duration of one
tracking window to avoid repeated activations/deactivations when psi
signal is bouncing.

Notifications to the users are rate-limited to one per tracking window.

Link: http://lkml.kernel.org/r/20190319235619.260832-8-surenb@google.com
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Dennis Zhou <dennis@kernel.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Li Zefan <lizefan@huawei.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

include/: refactor headers to allow kthread.h inclusion in psi_types.h

kthread.h can't be included in psi_types.h because it creates a circular
inclusion with kthread.h eventually including psi_types.h and
complaining on kthread structures not being defined because they are
defined further in the kthread.h.  Resolve this by removing psi_types.h
inclusion from the headers included from kthread.h.

Link: http://lkml.kernel.org/r/20190319235619.260832-7-surenb@google.com
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Dennis Zhou <dennis@kernel.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Li Zefan <lizefan@huawei.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

psi: track changed states

Introduce changed_states parameter into collect_percpu_times to track
the states changed since the last update.

This will be needed to detect whether polled states activated in the
monitor patch.

Link: http://lkml.kernel.org/r/20190319235619.260832-6-surenb@google.com
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Dennis Zhou <dennis@kernel.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Li Zefan <lizefan@huawei.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

psi: split update_stats into parts

Split update_stats into collect_percpu_times and update_averages for
collect_percpu_times to be reused later inside psi monitor.

Link: http://lkml.kernel.org/r/20190319235619.260832-5-surenb@google.com
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Dennis Zhou <dennis@kernel.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Li Zefan <lizefan@huawei.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

psi: rename psi fields in preparation for psi trigger addition

Rename psi_group structure member fields used for calculating psi totals
and averages for clear distinction between them and for trigger-related
fields that will be added by "psi: introduce psi monitor".

[surenb@google.com: v6]
Link: http://lkml.kernel.org/r/20190319235619.260832-4-surenb@google.com
Link: http://lkml.kernel.org/r/20190124211518.244221-5-surenb@google.com
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Dennis Zhou <dennis@kernel.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Li Zefan <lizefan@huawei.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

psi: make psi_enable static

psi_enable is not used outside of psi.c, make it static.

Link: http://lkml.kernel.org/r/20190319235619.260832-3-surenb@google.com
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Suggested-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

psi: introduce state_mask to represent stalled psi states

Patch series "psi: pressure stall monitors", v6.

This is a respin of:
  https://lwn.net/ml/linux-kernel/20190308184311.144521-1-surenb%40google.com/

Android is adopting psi to detect and remedy memory pressure that
results in stuttering and decreased responsiveness on mobile devices.

Psi gives us the stall information, but because we're dealing with
latencies in the millisecond range, periodically reading the pressure
files to detect stalls in a timely fashion is not feasible.  Psi also
doesn't aggregate its averages at a high-enough frequency right now.

This patch series extends the psi interface such that users can
configure sensitive latency thresholds and use poll() and friends to be
notified when these are breached.

As high-frequency aggregation is costly, it implements an aggregation
method that is optimized for fast, short-interval averaging, and makes
the aggregation frequency adaptive, such that high-frequency updates
only happen while monitored stall events are actively occurring.

With these patches applied, Android can monitor for, and ward off,
mounting memory shortages before they cause problems for the user.  For
example, using memory stall monitors in userspace low memory killer
daemon (lmkd) we can detect mounting pressure and kill less important
processes before device becomes visibly sluggish.  In our memory stress
testing psi memory monitors produce roughly 10x less false positives
compared to vmpressure signals.  Having ability to specify multiple
triggers for the same psi metric allows other parts of Android framework
to monitor memory state of the device and act accordingly.

The new interface is straight-forward.  The user opens one of the
pressure files for writing and writes a trigger description into the
file descriptor that defines the stall state - some or full, and the
maximum stall time over a given window of time.  E.g.:

        /* Signal when stall time exceeds 100ms of a 1s window */
        char trigger[] = "full 100000 1000000"
        fd = open("/proc/pressure/memory")
        write(fd, trigger, sizeof(trigger))
        while (poll() >= 0) {
                ...
        };
        close(fd);

When the monitored stall state is entered, psi adapts its aggregation
frequency according to what the configured time window requires in order
to emit event signals in a timely fashion.  Once the stalling subsides,
aggregation reverts back to normal.

The trigger is associated with the open file descriptor.  To stop
monitoring, the user only needs to close the file descriptor and the
trigger is discarded.

Patches 1-6 prepare the psi code for polling support.  Patch 7
implements the adaptive polling logic, the pressure growth detection
optimized for short intervals, and hooks up write() and poll() on the
pressure files.

The patches were developed in collaboration with Johannes Weiner.

This patch (of 7):

The psi monitoring patches will need to determine the same states as
record_times().  To avoid calculating them twice, maintain a state mask
that can be consulted cheaply.  Do this in a separate patch to keep the
churn in the main feature patch at a minimum.

This adds 4-byte state_mask member into psi_group_cpu struct which
results in its first cacheline-aligned part becoming 52 bytes long.  Add
explicit values to enumeration element counters that affect
psi_group_cpu struct size.

Link: http://lkml.kernel.org/r/20190124211518.244221-4-surenb@google.com
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Dennis Zhou <dennis@kernel.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Li Zefan <lizefan@huawei.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>