2
0
mirror of https://github.com/edk2-porting/linux-next.git synced 2024-12-23 12:43:55 +08:00
linux-next/tools
Andrii Nakryiko d41bc48bfa selftests/bpf: Add uprobe triggering overhead benchmarks
Add benchmark to measure overhead of uprobes and uretprobes. Also have
a baseline (no uprobe attached) benchmark.

On my dev machine, baseline benchmark can trigger 130M user_target()
invocations. When uprobe is attached, this falls to just 700K. With
uretprobe, we get down to 520K:

  $ sudo ./bench trig-uprobe-base -a
  Summary: hits  131.289 ± 2.872M/s

  # UPROBE
  $ sudo ./bench -a trig-uprobe-without-nop
  Summary: hits    0.729 ± 0.007M/s

  $ sudo ./bench -a trig-uprobe-with-nop
  Summary: hits    1.798 ± 0.017M/s

  # URETPROBE
  $ sudo ./bench -a trig-uretprobe-without-nop
  Summary: hits    0.508 ± 0.012M/s

  $ sudo ./bench -a trig-uretprobe-with-nop
  Summary: hits    0.883 ± 0.008M/s

So there is almost 2.5x performance difference between probing nop vs
non-nop instruction for entry uprobe. And 1.7x difference for uretprobe.

This means that non-nop uprobe overhead is around 1.4 microseconds for uprobe
and 2 microseconds for non-nop uretprobe.

For nop variants, uprobe and uretprobe overhead is down to 0.556 and
1.13 microseconds, respectively.

For comparison, just doing a very low-overhead syscall (with no BPF
programs attached anywhere) gives:

  $ sudo ./bench trig-base -a
  Summary: hits    4.830 ± 0.036M/s

So uprobes are about 2.67x slower than pure context switch.

Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20211116013041.4072571-1-andrii@kernel.org
2021-11-16 14:46:49 +01:00
..
accounting
arch Merge remote-tracking branch 'torvalds/master' into perf/core 2021-11-06 15:49:33 -03:00
bootconfig bootconfig: Cleanup dummy headers in tools/bootconfig 2021-10-10 22:16:02 -04:00
bpf bpftool: Update doc (use susbtitutions) and test_bpftool_synctypes.py 2021-11-16 13:56:22 +01:00
build tools: Bump minimum LLVM C++ std to GNU++14 2021-11-04 09:31:30 -03:00
cgroup
counter tools/counter: Create Counter tools 2021-10-17 10:54:16 +01:00
debugging tools: Fix "the the" in a message in kernel-chktaint 2021-06-13 17:01:17 -06:00
edid
firewire
firmware
gpio
hv
iio
include bpf: Change value of MAX_TAIL_CALL_CNT from 32 to 33 2021-11-16 14:03:15 +01:00
io_uring tools/io_uring/io_uring-cp: sync with liburing example 2021-08-13 08:58:11 -06:00
kvm/kvm_stat KVM: kvm_stat: do not show halt_wait_ns 2021-10-18 14:07:18 -04:00
laptop
leds
lib libbpf: Fix a couple of missed btf_type_tag handling in btf.c 2021-11-16 13:10:52 +01:00
memory-model tools/memory-model: Document data_race(READ_ONCE()) 2021-07-27 11:48:55 -07:00
objtool Tracing updates for 5.16: 2021-11-01 20:05:19 -07:00
pci tools: PCI: Zero-initialize param 2021-08-05 11:01:30 +01:00
pcmcia
perf Add 'tools/perf/libbpf/' to ignored files 2021-11-08 11:33:35 -08:00
power tools/power/x86/intel-speed-select: v1.10 release 2021-06-18 15:29:32 +02:00
rcu tools/rcu: Add an extract-stall script 2021-09-16 10:31:26 -07:00
scripts tools, build: Add RISC-V to HOSTARCH parsing 2021-11-01 17:08:21 +01:00
spi
testing selftests/bpf: Add uprobe triggering overhead benchmarks 2021-11-16 14:46:49 +01:00
thermal/tmon tools/thermal/tmon: Add cross compiling support 2021-08-14 15:33:19 +02:00
time
tracing tools/latency-collector: Use correct size when writing queue_full_warning 2021-10-25 22:27:19 -04:00
usb usb: testusb: Fix for showing the connection speed 2021-09-14 10:31:41 +02:00
virtio tools/virtio: fix build 2021-08-11 06:44:24 -04:00
vm tools/vm/page-types.c: print file offset in hexadecimal 2021-11-06 13:30:40 -07:00
wmi
Makefile tools/lib/lockdep: drop liblockdep 2021-11-12 11:07:17 -08:00