Lima: Linux-on-Mac ("macOS subsystem for Linux", "containerd for Mac")

Lima: Linux-on-Mac ("macOS subsystem for Linux", "containerd for Mac")

Lima launches Linux virtual machines on macOS, with automatic file sharing, port forwarding, and containerd.

Lima can be considered as a some sort of unofficial "macOS subsystem for Linux", or "containerd for Mac".

✅ Automatic file sharing

✅ Automatic port forwarding

✅ Built-in support for containerd

✅ Intel on Intel

✅ ARM on Intel

✅ ARM on ARM (untested)

✅ Intel on ARM (untested)

✅ Ubuntu

✅ Fedora

Related project: sshocker (ssh with file sharing and port forwarding)

Examples

uname

$ uname -a
Darwin macbook.local 20.4.0 Darwin Kernel Version 20.4.0: Thu Apr 22 21:46:47 PDT 2021; root:xnu-7195.101.2~1/RELEASE_X86_64 x86_64

$ lima uname -a
Linux lima-default 5.11.0-16-generic #17-Ubuntu SMP Wed Apr 14 20:12:43 UTC 2021 x86_64 x86_64 x86_64 GNU/Linux

$ LIMA_INSTANCE=arm lima uname -a
Linux lima-arm 5.11.0-16-generic #17-Ubuntu SMP Wed Apr 14 20:10:16 UTC 2021 aarch64 aarch64 aarch64 GNU/Linux

Sharing files across macOS and Linux

$ echo "files under /Users on macOS filesystem are readable from Linux" > some-file

$ lima cat some-file
files under /Users on macOS filesystem are readable from Linux

$ lima sh -c 'echo "/tmp/lima is writable from both macOS and Linux" > /tmp/lima/another-file'

$ cat /tmp/lima/another-file
/tmp/lima is writable from both macOS and Linux"

Running containerd containers (compatible with Docker containers)

$ lima nerdctl run -d --name nginx -p 127.0.0.1:8080:80 nginx:alpine

http://127.0.0.1:8080 is accessible from both macOS and Linux.

NOTE Privileged ports (0-1023) cannot be forwarded

For the usage of containerd and nerdctl (contaiNERD ctl), visit https://github.com/containerd/containerd and https://github.com/containerd/nerdctl.

Getting started

Requirements (Intel Mac)

• coreutils (for realpath command) (brew install coreutils)

• QEMU (brew install qemu)

• Run the following commands to enable --accel=hvf:

cat >entitlements.xml <<EOF
<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
<dict>
    <!-- for OS X 10.10 - macOS 10.15 -->
    <key>com.apple.vm.hypervisor</key>
    <true/>
    <!-- for macOS 11 and later -->
    <key>com.apple.security.hypervisor</key>
    <true/>
</dict>
</plist>
EOF

codesign -s - --entitlements entitlements.xml --force /usr/local/bin/qemu-system-x86_64

Requirements (ARM Mac)

• coreutils (for realpath command) (brew install coreutils)
• QEMU with --accel=hvf support, see https://gist.github.com/citruz/9896cd6fb63288ac95f81716756cb9aa

NOTE Lima is not tested on ARM Mac.

Install

Download the binary archive from https://github.com/AkihiroSuda/lima/releases , and extract it under /usr/local (or somewhere else).

To install from the source, run make && make install.

Usage

• Run limactl start <INSTANCE> to start the Linux instance. The default instance name is "default". Lima automatically opens an editor (vi) for reviewing and modifying the configuration. Wait until "READY" to be printed on the host terminal.

• Run limactl shell <INSTANCE> <COMMAND> to launch <COMMAND> on Linux. For the "default" instance, this command can be shortened as just lima <COMMAND>. The lima command also accepts the instance name as the environment variable $LIMA_INSTANCE.

• Run limactl ls to show the instances.

• Run limactl delete <INSTANCE> to delete the instance.

• To enable bash completion, add source <(limactl completion bash) to ~/.bash_profile.

⚠️ CAUTION: make sure to back up your data

Lima may have bugs that result in loss of data.

Make sure to back up your data before running Lima.

Especially, the following data might be easily lost:

• Data in the shared writable directories (/tmp/lima by default), probably after hibernation of the host machine (e.g., after closing and reopening the laptop lid)
• Data in the VM image, mostly when upgrading the version of lima

Configuration

See ./pkg/limayaml/default.TEMPLATE.yaml.

The current default spec:

• OS: Ubuntu 21.04 (Hirsute Hippo)
• CPU (x86_64): Haswell v4, 4 cores
• CPU (aarch64): Cortex A72, 4 cores
• Memory: 4 GiB
• Disk: 100 GiB
• Mounts: ~ (read-only), /tmp/lima (writable)
• SSH: 127.0.0.1:60022

How it works

• Hypervisor: QEMU with HVF accelerator
• Filesystem sharing: reverse sshfs
• Port forwarding: ssh -L, automated by watching /proc/net/tcp in the guest

Developer guide

Contributing to Lima

• Please certify your Developer Certificate of Origin (DCO), by signing off your commit with git commit -s and with your real name.

• Please squash commits.

Help wanted

🙏

• Test on ARM Mac
• Performance optimization
• Homebrew
• More guest distros
• Linux on Windows/Linux/BSD
• GUI with system tray icon (Qt or Electron, for portability)
• VirtFS to replace the current reverse sshfs (work has to be done on QEMU repo)
• vsock to replace SSH (work has to be done on QEMU repo)

FAQs

"What's my login password?"

Password is disabled and locked by default. You have to use limactl shell bash (or lima bash) to open a shell.

Alternatively, you may also directly ssh into the guest: ssh -p 60022 -o NoHostAuthenticationForLocalhost=yes 127.0.0.1.

"Does Lima work on ARM Mac?"

Yes, it should work, but not tested on ARM.

"Can I run non-Ubuntu guests?"

Fedora is also known to work, see ./examples/fedora.yaml. This file can be loaded with limactl start ./examples/fedora.yaml.

An image has to satisfy the following requirements:

• systemd
• cloud-init
• The following binaries to be preinstalled:
  • curl
  • sudo
• The following binaries to be preinstalled, or installable via the package manager:
  • sshfs
  • newuidmap and newgidmap
• apt-get or dnf (if you want to contribute support for another package manager, run git grep apt-get to find out where to modify)

"Can I run other container engines such as Podman?"

Yes, if you install it.

containerd can be stopped with systemctl --user disable --now containerd.

"Can I run Lima with a remote Linux machine?"

Lima itself does not support connecting to a remote Linux machine, but sshocker, the predecessor or Lima, provides similar features for remote Linux machines.

e.g., run sshocker -v /Users/foo:/home/foo/mnt -p 8080:80 <USER>@<HOST> to expose /Users/foo to the remote machine as /home/foo/mnt, and forward localhost:8080 to the port 80 of the remote machine.

"QEMU crashes with HV_ERROR"

You have to add com.apple.security.hypervisor entitlement to qemu-system-x86_64 binary. See Getting started.

"QEMU is slow"

• Make sure that HVF is enabled with com.apple.security.hypervisor entitlement. See Getting started.
• Emulating non-native machines (ARM-on-Intel, Intel-on-ARM) is slow by design.

"Port forwarding does not work"

Privileged ports (0-1023) cannot be forwarded. e.g., you have to use 8080, not 80.

error "field SSHPubKeys must be set"

Make sure you have a ssh keypair in ~/.ssh. To create:

ssh-keygen -q -t rsa -N '' -f ~/.ssh/id_rsa <<<n 2>&1 >/dev/null

error "hostkeys_foreach failed: No such file or directory"

Make sure you have a ssh known_hosts file:

touch ~/.ssh/known_hosts

error "failed to execute script ssh: [...] Permission denied (publickey)"

If you have a ~/.ssh/config with a username overwrite for all hosts, exclude 127.0.0.1 from it. Example:

Host * !127.0.0.1
        User root

"Hints for debugging other problems?"

• Inspect logs:
  • limactl --debug start
  • /var/log/cloud-init-output.log (inside the guest)
  • /var/log/cloud-init.log (inside the guest)
• Make sure that you aren't mixing up tabs and spaces in the YAML.
• If you have passphrases for any private key under ~/.ssh, you will need to have ssh-agent running.