nerves.system.shell fails when Docker VirtioFS is enabled

[tonnenpinguin]

Environment

• Elixir version (elixir -v):

Erlang/OTP 24 [erts-12.0.3] [source] [64-bit] [smp:12:12] [ds:12:12:10] [async-threads:1] [jit]

Elixir 1.13.3 (compiled with Erlang/OTP 24)

• Nerves environment: (mix nerves.env --info)

|nerves_bootstrap| Environment Package List

  Pkg:         nerves_system_br
  Vsn:         1.18.6
  Type:        system_platform
  BuildRunner: {nil, []}

  Pkg:         nerves_system_rpi4
  Vsn:         1.18.4
  Type:        system
  BuildRunner: {Nerves.Artifact.BuildRunners.Docker, [make_args: ["source", "all", "legal-info"]]}

  Pkg:         nerves_toolchain_aarch64_nerves_linux_gnu
  Vsn:         1.5.0
  Type:        toolchain
  BuildRunner: {Nerves.Artifact.BuildRunners.Local, []}

  Pkg:         nerves_toolchain_ctng
  Vsn:         1.8.5
  Type:        toolchain_platform
  BuildRunner: {nil, []}

|nerves_bootstrap| Loadpaths Start


Nerves environment
  MIX_TARGET:   target
  MIX_ENV:      dev

|nerves_bootstrap| Environment Variable List
  target:     target
  toolchain:  /Users/bernhardhackl/.nerves/artifacts/nerves_toolchain_aarch64_nerves_linux_gnu-darwin_x86_64-1.5.0
  system:     /Users/bernhardhackl/.nerves/artifacts/nerves_system_rpi4-portable-1.18.4
  app:        .

|nerves_bootstrap| Loadpaths End

• Additional information about your host, target hardware or environment that
  may help

Docker Desktop 4.7.0 (77141) with VirtioFS enabled
macOS Monterey 12.3 (Intel chipset)

Current behavior

When running mix nerves.system.shell I get the following error.

➜  nerves_system_custom_rpi4 git:(434cae6) ✗ mix nerves.system.shell
==> nerves
stty sane rows 30 cols 187; stty -echo
export PS1=""; export PS2=""
start() {
echo -e "\e[25F\e[0J\e[1;7m
  Preparing Nerves Shell  \e[0m"
echo -e "\e]0;Nerves Shell\a"
export PS1="\e[1;7m Nerves \e[0;1m \W > \e[0m"
export PS2="\e[1;7m Nerves \e[0;1m \W ..\e[0m"
echo Updating build directory.
echo This will take a while if it is the first time...
/nerves/env/platform/create-build.sh /nerves/env/nerves_system_rpi4/nerves_defconfig /nerves/build >/dev/null
stty echo
}; start
docker: Error response from daemon: error while creating mount source path '/private/tmp/com.apple.launchd.roIUXGKQgd/Listeners': mkdir /private/tmp/com.apple.launchd.roIUXGKQgd/Listeners: operation not supported.

I found this issue that suggests to set SSH_AUTH_SOCK to /run/host-services/ssh-auth.sock.

If I run export SSH_AUTH_SOCK=/run/host-services/ssh-auth.sock before mix nerves.system.shell docker at least seems to start properly.

Now I get the following errors when trying to untar the buildroot. That error should already be fixed in the upcoming Docker release though.

tar: board/olimex/a10_olinuxino/boot.cmd: Cannot open: Permission denied
tar: board/olimex/a10_olinuxino/genimage.cfg: Cannot open: Permission denied
tar: board/olimex/a10_olinuxino/post-build.sh: Cannot open: Permission denied
tar: board/solidrun/clearfog_gt_8k/genimage.cfg: Cannot open: Permission denied
tar: board/solidrun/clearfog_gt_8k/post-build-mainline.sh: Cannot open: Permission denied
tar: board/solidrun/clearfog_gt_8k/uboot-fragment.config: Cannot open: Permission denied
tar: boot/barebox/barebox-aux/barebox-aux.hash: Cannot open: Permission denied
tar: boot/barebox/barebox/barebox.hash: Cannot open: Permission denied
tar: package/bluez5_utils-headers/bluez5_utils-headers.hash: Cannot open: Permission denied
tar: package/busybox/S02sysctl: Cannot open: Permission denied
tar: package/fftw/fftw-double/fftw-double.hash: Cannot open: Permission denied
tar: package/fftw/fftw-long-double/fftw-long-double.hash: Cannot open: Permission denied
tar: package/fftw/fftw-quad/fftw-quad.hash: Cannot open: Permission denied
[...]

If I disable VirtioFS everything works as expected though :)

Expected behavior

To support Docker VirtioFS out of the box :)

[fhunleth]

Thanks for the really descriptive report! This is really outside my expertise, so I'm hoping for some help. If no one posts that they can help in the next couple weeks, I'll see what I can figure out.