Rosetta in Linux

2023年 10月 17日火曜日

これは 2023/10/17 の MK 社内 LT で発表した内容です。

目的 (Goal)

Apple シリコン上の ARM Linux で x86_64 Linux バイナリを実行する
Running x86_64 Linux binaries on ARM Linux on Apple Silicon
公式開発ドキュメント：Running Intel Binaries in Linux VMs with Rosetta¹

実現方法 (Method)

Kernel Support for miscellaneous Binary Formats (binfmt_misc)²

binfmt_misc

カーネル機能 (Kernel function)
インタプリタさえあればどんなバイナリでも実行できる
Any binary can be executed as long as there is an interpreter

原理図解 (Principle Diagram)

ホスト CPU アーキテクチャ用のプログラムを実行する
Run a program for the host CPU architecture
他の CPU アーキテクチャのプログラムを実行する
Run a program for the other CPU architecture

設定例 (Settings Example)

Rosetta

% mkdir /tmp/mountpoint
% sudo mount -t virtiofs EXAMPLE_TAG /tmp/mountpoint
% ls /tmp/mountpoint rosetta
% sudo /usr/sbin/update-binfmts --install rosetta /tmp/mountpoint/rosetta \
    --magic "\x7fELF\x02\x01\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x02\x00\x3e\x00" \
    --mask "\xff\xff\xff\xff\xff\xfe\xfe\x00\xff\xff\xff\xff\xff\xff\xff\xff\xfe\xff\xff\xff" \
    --credentials yes --preserve yes --fix-binary yes

QEMU (arm64)

echo ':arm:M::\x7fELF\x01\x01\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x02\x00\x28\x00:\xff\xff\xff\xff\xff\xff\xff\x00\xff\xff\xff\xff\xff\xff\xff\xff\xfe\xff\xff\xff:/usr/bin/qemu-arm64-static:' > /proc/sys/fs/binfmt_misc/register

設定結果を見る (Check the settings)

cliff@PD-Jammy-00:~$ ls /proc/sys/fs/binfmt_misc
python3.10  register  RosettaLinux  status

cliff@PD-Jammy-00:~$ cat /proc/sys/fs/binfmt_misc/RosettaLinux
enabled
interpreter /media/psf/RosettaLinux/rosetta
flags: OCF
offset 0
magic 7f454c4602010100000000000000000002003e00
mask fffffffffffefe00fffffffffffffffffeffffff

動作環境 (Requirement)

macOS 13+
UTM or Parallels Desktop with ARM64 Linux VM

使い方 (Usage)

UTM

UTM / Advanced / Rosetta

Parallels Desktop

Using Rosetta to run x86-64 Docker Containers and Binaries in Linux virtual machines with Parallels Desktop

ベンチマーク (Benchmark)

sudo apt install sysbench
sysbench cpu run

ARM64: Native (Virtualization by Parallels Desktop)

cliff@PD-Jammy-00:~$ sysbench cpu run
sysbench 1.0.20 (using system LuaJIT 2.1.0-beta3)

Running the test with following options:
Number of threads: 1
Initializing random number generator from current time


Prime numbers limit: 10000

Initializing worker threads...

Threads started!

CPU speed:
    events per second:  8425.79

General statistics:
    total time:                          10.0002s
    total number of events:              84266

Latency (ms):
         min:                                    0.11
         avg:                                    0.12
         max:                                    0.23
         95th percentile:                        0.12
         sum:                                 9982.57

Threads fairness:
    events (avg/stddev):           84266.0000/0.00
    execution time (avg/stddev):   9.9826/0.00

ARM64: Docker (Native)

docker@PD-Jammy-00:~$ sysbench cpu run
sysbench 1.0.20 (using system LuaJIT 2.1.0-beta3)

Running the test with following options:
Number of threads: 1
Initializing random number generator from current time


Prime numbers limit: 10000

Initializing worker threads...

Threads started!

CPU speed:
    events per second:  8310.11

General statistics:
    total time:                          10.0001s
    total number of events:              83108

Latency (ms):
         min:                                    0.11
         avg:                                    0.12
         max:                                    0.26
         95th percentile:                        0.13
         sum:                                 9974.35

Threads fairness:
    events (avg/stddev):           83108.0000/0.00
    execution time (avg/stddev):   9.9743/0.00

x86_64: Docker (QEMU)³⁴

docker run --privileged --rm tonistiigi/binfmt --install amd64
/usr/bin/gemu-×86_64 /usr/bin/sysbench sysbench cpu run

docker@PD-Jammy-00:~$ sysbench cpu run
sysbench 1.0.20 (using system LuaJIT 2.1.0-beta3)

Running the test with following options:
Number of threads: 1
Initializing random number generator from current time


Prime numbers limit: 10000

Initializing worker threads...

Threads started!

CPU speed:
    events per second:   283.95

General statistics:
    total time:                          10.0036s
    total number of events:              2841

Latency (ms):
         min:                                    3.35
         avg:                                    3.52
         max:                                    8.49
         95th percentile:                        3.75
         sum:                                 9993.88

Threads fairness:
    events (avg/stddev):           2841.0000/0.00
    execution time (avg/stddev):   9.9939/0.00

x86_64: Docker (Rosetta)

/media/psf/RosettaLinux/rosetta /usr/bin/sysbench cpu run

docker@PD-Jammy-00:~$ sysbench cpu run
sysbench 1.0.20 (using system LuaJIT 2.1.0-beta3)

Running the test with following options:
Number of threads: 1
Initializing random number generator from current time


Prime numbers limit: 10000

Initializing worker threads...

Threads started!

CPU speed:
    events per second:  7000.22

General statistics:
    total time:                          10.0006s
    total number of events:              70012

Latency (ms):
         min:                                    0.14
         avg:                                    0.14
         max:                                    0.80
         95th percentile:                        0.15
         sum:                                 9976.68

Threads fairness:
    events (avg/stddev):           70012.0000/0.00
    execution time (avg/stddev):   9.9767/0.00

結果 (Result)

試したプログラム (Tried and Tested Programs)

Google Chrome

普通に使えると思うですが、YouTube の再生不能
I think it can be used normally, but YouTube cannot be played

ALSA lib conf.c:4002:(snd_config_hooks_call) Cannot open shared library libasound_module_conf_pulse.so (/usr/lib/x86_64-linux-gnu/alsa-lib/libasound_module_conf_pulse.so: cannot open shared object file: No such file or directory)
ALSA lib control.c:1528:(snd_ctl_open_noupdate) Invalid CTL hw:0

NVIDIA SDK Manager

Jetson のイメージは作成できるが、焼くことはできない (NVIDIA SDK Manager からの制限)
Jetson image can be created, but cannot be burned (limitation from NVIDIA SDK Manager)

Reference

この記事をシェア

目的 (Goal)

実現方法 (Method)

binfmt_misc

原理図解 (Principle Diagram)

設定例 (Settings Example)

設定結果を見る (Check the settings)

動作環境 (Requirement)

使い方 (Usage)

UTM

Parallels Desktop

ベンチマーク (Benchmark)

ARM64: Native (Virtualization by Parallels Desktop)

ARM64: Docker (Native)

x86_64: Docker (QEMU)34

x86_64: Docker (Rosetta)

結果 (Result)

試したプログラム (Tried and Tested Programs)

Google Chrome

NVIDIA SDK Manager

Reference

x86_64: Docker (QEMU)³⁴