linux_dsm_epyc7002/arch
Yonghong Song e7ed9d9bd0 uprobes/x86: Emulate push insns for uprobe on x86
Uprobe is a tracing mechanism for userspace programs.
Typical uprobe will incur overhead of two traps.
First trap is caused by replaced trap insn, and
the second trap is to execute the original displaced
insn in user space.

To reduce the overhead, kernel provides hooks
for architectures to emulate the original insn
and skip the second trap. In x86, emulation
is done for certain branch insns.

This patch extends the emulation to "push <reg>"
insns. These insns are typical in the beginning
of the function. For example, bcc
in https://github.com/iovisor/bcc repo provides
tools to measure funclantency, detect memleak, etc.
The tools will place uprobes in the beginning of
function and possibly uretprobes at the end of function.
This patch is able to reduce the trap overhead for
uprobe from 2 to 1.

Without this patch, uretprobe will typically incur
three traps. With this patch, if the function starts
with "push" insn, the number of traps can be
reduced from 3 to 2.

An experiment was conducted on two local VMs,
fedora 26 64-bit VM and 32-bit VM, both 4 processors
and 4GB memory, booted with latest tip repo (and this patch).
The host is MacBook with intel i7 processor.

The test program looks like:

  #include <stdio.h>
  #include <stdlib.h>
  #include <time.h>
  #include <sys/time.h>

  static void test() __attribute__((noinline));
  void test() {}
  int main() {
    struct timeval start, end;

    gettimeofday(&start, NULL);
    for (int i = 0; i < 1000000; i++) {
      test();
    }
    gettimeofday(&end, NULL);

    printf("%ld\n", ((end.tv_sec * 1000000 + end.tv_usec)
                     - (start.tv_sec * 1000000 + start.tv_usec)));
    return 0;
  }

The program is compiled without optimization, and
the first insn for function "test" is "push %rbp".
The host is relatively idle.

Before the test run, the uprobe is inserted as below for uprobe:
  echo 'p <binary>:<test_func_offset>' > /sys/kernel/debug/tracing/uprobe_events
  echo 1 > /sys/kernel/debug/tracing/events/uprobes/enable
and for uretprobe:
  echo 'r <binary>:<test_func_offset>' > /sys/kernel/debug/tracing/uprobe_events
  echo 1 > /sys/kernel/debug/tracing/events/uprobes/enable

Unit: microsecond(usec) per loop iteration

x86_64          W/ this patch   W/O this patch
uprobe          1.55            3.1
uretprobe       2.0             3.6

x86_32          W/ this patch   W/O this patch
uprobe          1.41            3.5
uretprobe       1.75            4.0

You can see that this patch significantly reduced the overhead,
50% for uprobe and 44% for uretprobe on x86_64, and even more
on x86_32.

Signed-off-by: Yonghong Song <yhs@fb.com>
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: kernel-team@fb.com
Link: http://lkml.kernel.org/r/20171201001202.3706564-1-yhs@fb.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-12-11 18:42:11 +01:00
..
alpha treewide: setup_timer() -> timer_setup() 2017-11-21 15:57:07 -08:00
arc ARC updates for 4.15-rc1 2017-11-25 08:21:54 -10:00
arm Merge branch 'fixes' of git://git.armlinux.org.uk/~rmk/linux-arm 2017-12-03 10:51:08 -05:00
arm64 arm64 fixes: 2017-12-01 19:37:03 -05:00
blackfin treewide: setup_timer() -> timer_setup() 2017-11-21 15:57:07 -08:00
c6x Kbuild updates for v4.15 2017-11-17 17:45:29 -08:00
cris pci-v4.15-changes 2017-11-15 15:01:28 -08:00
frv Kbuild updates for v4.15 2017-11-17 17:45:29 -08:00
h8300 mm, arch: remove empty_bad_page* 2017-11-15 18:21:03 -08:00
hexagon Kbuild updates for v4.15 2017-11-17 17:45:29 -08:00
ia64 arch/ia64/include/asm/topology.h: remove unused parent_node() macro 2017-11-17 16:10:04 -08:00
m32r m32r: fix endianness constraints 2017-11-15 18:21:00 -08:00
m68k m68k/macboing: Fix missed timer callback assignment 2017-11-24 16:19:40 +01:00
metag DeviceTree for 4.15: 2017-11-14 18:25:40 -08:00
microblaze Microblaze patch for 4.15-rc2 2017-11-29 14:19:22 -08:00
mips TTY/Serial driver fixes for 4.15-rc3 2017-12-05 09:05:16 -08:00
mn10300 bug: define the "cut here" string in a single place 2017-11-17 16:10:01 -08:00
nios2 DeviceTree for 4.15: 2017-11-14 18:25:40 -08:00
openrisc kmemcheck: remove annotations 2017-11-15 18:21:04 -08:00
parisc treewide: Switch DEFINE_TIMER callbacks to struct timer_list * 2017-11-21 15:57:05 -08:00
powerpc powerpc fixes for 4.15 2017-12-01 08:40:17 -05:00
riscv RISC-V: Fixes for clean allmodconfig build 2017-12-01 13:31:31 -08:00
s390 * x86 bugfixes: APIC, nested virtualization, IOAPIC 2017-11-30 08:15:19 -08:00
score License cleanup: add SPDX license identifier to uapi header files with no license 2017-11-02 11:19:54 +01:00
sh treewide: setup_timer() -> timer_setup() 2017-11-21 15:57:07 -08:00
sparc Merge branch 'akpm' (patches from Andrew) 2017-11-29 19:12:44 -08:00
tile mm: switch to 'define pmd_write' instead of __HAVE_ARCH_PMD_WRITE 2017-11-29 18:40:42 -08:00
um This pull request contains the following core changes: 2017-11-22 20:46:06 -10:00
unicore32 kmemcheck: stop using GFP_NOTRACK and SLAB_NOTRACK 2017-11-15 18:21:04 -08:00
x86 uprobes/x86: Emulate push insns for uprobe on x86 2017-12-11 18:42:11 +01:00
xtensa libnvdimm for 4.15 2017-11-17 09:51:57 -08:00
.gitignore
Kconfig bpf: Revert bpf_overrid_function() helper changes. 2017-11-11 18:24:55 +09:00