podman create/run --shm-size doesn't work

[debarshiray]

/kind bug

Description

While investigating why /dev/shm wasn't a tmpfs, I found that podman create --shm-size or podman run --shm-size doesn't seem to work as advertised.

$ podman run --interactive --name foo --shm-size 1g --tty fedora /bin/sh
sh-4.4# df -h /dev/shm
Filesystem      Size  Used Avail Use% Mounted on
/dev/nvme0n1p5  415G   26G  368G   7% /dev/shm

Notice how the size of /dev/shm isn't 1G. In fact, it doesn't have any effect, because the result is the same without --shm-size.

$ podman run --interactive --name foo --tty fedora /bin/sh
sh-4.4# df -h /dev/shm
Filesystem      Size  Used Avail Use% Mounted on
/dev/nvme0n1p5  415G   26G  368G   7% /dev/shm

I get similar results with podman create, which is what the Silverblue toolbox uses.

(Why doesn't podman mount a tmpfs at /dev/shm by default?)

Output of podman version:

Version:       0.10.1.3
Go Version:    go1.10.4
OS/Arch:       linux/amd64

Output of podman info:

host:
  BuildahVersion: 1.5-dev
  Conmon:
    package: podman-0.10.1.3-3.gitdb08685.fc28.x86_64
    path: /usr/libexec/podman/conmon
    version: 'conmon version 1.12.0-dev, commit: 7ab6aac44ec489c04baa7555107ffe668b210b96-dirty'
  Distribution:
    distribution: fedora
    version: "28"
  MemFree: 4783382528
  MemTotal: 16696295424
  OCIRuntime:
    package: runc-1.0.0-57.dev.git9e5aa74.fc28.x86_64
    path: /usr/bin/runc
    version: 'runc version spec: 1.0.1-dev'
  SwapFree: 4208979968
  SwapTotal: 4208979968
  arch: amd64
  cpus: 4
  hostname: bollard
  kernel: 4.18.16-200.fc28.x86_64
  os: linux
  uptime: 1h 22m 47.86s (Approximately 0.04 days)
insecure registries:
  registries: []
registries:
  registries:
  - docker.io
  - registry.fedoraproject.org
  - quay.io
  - registry.access.redhat.com
  - registry.centos.org
store:
  ContainerStore:
    number: 1
  GraphDriverName: vfs
  GraphOptions: []
  GraphRoot: /var/home/rishi/.local/share/containers/storage
  GraphStatus: {}
  ImageStore:
    number: 5
  RunRoot: /run/user/1000/run

Additional environment details (AWS, VirtualBox, physical, etc.):

This is a physical laptop running Fedora 28 Silverblue 28.20181105.0, with podman-0.10.1.3-3.gitdb08685.fc28 and 2:runc-1.0.0-57.dev.git9e5aa74.fc28 overlaid on top.

[rhatdan]

#  podman run --interactive --name foo1 --tty fedora mount | grep shm
shm on /dev/shm type tmpfs (rw,nosuid,nodev,noexec,relatime,context="system_u:object_r:container_file_t:s0:c162,c460",size=64000k)

shm-size is used to modify the default size setting in the mount option. By default we mount at 64000k

#  podman run --interactive --name foo --shm-size 1g --tty fedora mount | grep shm
shm on /dev/shm type tmpfs (rw,nosuid,nodev,noexec,relatime,context="system_u:object_r:container_file_t:s0:c761,c1021",size=976564k)

Notice the mount point changed to 976564k

[mheon]

This looks like a rootless thing, but I'm confused as to why, we should be able to make tmpfs mounts in the rootless userns

…

On Wed, Nov 7, 2018, 08:00 Daniel J Walsh ***@***.*** wrote: # podman run --interactive --name foo1 --tty fedora mount | grep shm shm on /dev/shm type tmpfs (rw,nosuid,nodev,noexec,relatime,context="system_u:object_r:container_file_t:s0:c162,c460",size=64000k) shm-size is used to modify the default size setting in the mount option. By default we mount at 64000k # podman run --interactive --name foo --shm-size 1g --tty fedora mount | grep shm shm on /dev/shm type tmpfs (rw,nosuid,nodev,noexec,relatime,context="system_u:object_r:container_file_t:s0:c761,c1021",size=976564k) Notice the mount point changed to 976564k — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#1770 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AHYHCMvKc3Z5qE9wKiiEbCg-i_CLW0Haks5ustl-gaJpZM4YSWaH> .

[rhatdan]

@mheon User NS Allows you to mount tmpfs as long as it is on a directory that you own. But maybe it does not allow you to set the size.

[rhatdan]

@giuseppe Are we not using tmpfs mounted on /dev/shm inside of rootless containers?

[debarshiray]

Yeah, this is when running rootless.

It's not the end of the world because I can work around it with --mount type=tmpfs,destination=/dev/shm.

It has been trickier to mimic the in-kernel tmpfs default of using 50% of RAM instead of 64M. Let's hope my awk incantation to parse /proc/meminfo holds up. :)

[rhatdan]

I do think this is a bug in podman, we should be using a tmpfs for this not physical disk.

[giuseppe]

there is a PR for this issue: #1777

[rhatdan]

Since this PR has been merged, I am going to close this issue.