OverlayFS Finally Offered For Pulling Into Linux 3.18

[phoronix]

Phoronix: OverlayFS Finally Offered For Pulling Into Linux 3.18

When Linux 3.18-rc1 was released last week, one week sooner than anticipated, Linus Torvalds mentioned he was willing to still allow OverlayFS to be merged this cycle. One week later, that code is hopefully now ready for merging...

http://www.phoronix.com/vr.php?view=MTgyMjg

[q2dg]

How to know who is right?

People who says OverlayFS is "incomplete (or "is not happy with the design") or people who has merged it into mainline?
Who I trust?

[Zajec]

q2dg: you can check Okajima's reply: https://lkml.org/lkml/2014/9/30/35
There are some issues that according to the Miklos's reply (https://lkml.org/lkml/2014/9/30/107) are at least documented.

My guess is that you simply have to use "overlayfs" & be aware of it's behavior (see documentation).

[Gusar]

People who says OverlayFS is "incomplete (or "is not happy with the design") or people who has merged it into mainline?
Who I trust?

People *were* unhappy with the design, in particular Al Viro didn't like that OverlayFS used extended attributes for whiteouts. So Miklos changed the design to what Linus Torvalds proposed, to use a special character device for whiteouts. Now Al Viro is fine with OverlayFS and so it's being merged.

There are certain things OverlayFS doesn't handle, but these things aren't crucial in practice for the common uses cases and they're well documented in the kernel docs (pay particular attention to the "Non-standard behavior" part). Where these things do matter, AUFS should be used. AUFS offers tons more features, but this comes at a cost - it's a lot more complex and is almost ten times as big as OverlayFS.

[kirk]

Really was hoping AUFS would get merged. I wonder what size OverlayFS will be when it's usable for things beyond a live CD?

[droidhacker]

Really was hoping AUFS would get merged. I wonder what size OverlayFS will be when it's usable for things beyond a live CD?

It *IS* usable for things beyond a livecd.
Unless you happen to be aware of significant filesystem related problems coming from OpenWRT or its derivatives... OpenWRT uses overlayfs on *root* to transparently merge a writable jffs2 with a readonly squashfs.

[jonnor]

OverlayFS on root and similar hacks will hopefully no longer be neccesary, with the changes being pushed for stateless and factory-resettable systems being pushed forward by systemd project.

[caligula]

OverlayFS on root and similar hacks will hopefully no longer be neccesary, with the changes being pushed for stateless and factory-resettable systems being pushed forward by systemd project.

Are you troll or something? Do you have any idea how bloated systemd is and how little space there is on those boxes. The ROM could be 8 MB. This is from my x86-64 system:

9.5M /lib/systemd/
0 /lib/libsystemd-daemon.so
0 /lib/libsystemd-daemon.so.0
24K /lib/libsystemd-daemon.so.0.0.12
0 /lib/libsystemd-id128.so
0 /lib/libsystemd-id128.so.0
16K /lib/libsystemd-id128.so.0.0.28
0 /lib/libsystemd-journal.so
0 /lib/libsystemd-journal.so.0
116K /lib/libsystemd-journal.so.0.11.5
0 /lib/libsystemd-login.so
0 /lib/libsystemd-login.so.0
52K /lib/libsystemd-login.so.0.9.3
0 /lib/libsystemd.so
0 /lib/libsystemd.so.0
144K /lib/libsystemd.so.0.5.0
1.3M /sbin/systemd-analyze
44K /sbin/systemd-ask-password
32K /sbin/systemd-cat
252K /sbin/systemd-cgls
56K /sbin/systemd-cgtop
52K /sbin/systemd-delta
32K /sbin/systemd-detect-virt
36K /sbin/systemd-escape
72K /sbin/systemd-firstboot
240K /sbin/systemd-inhibit
36K /sbin/systemd-machine-id-setup
32K /sbin/systemd-notify
372K /sbin/systemd-nspawn
40K /sbin/systemd-path
260K /sbin/systemd-run
0 /sbin/systemd-stdio-bridge
76K /sbin/systemd-sysusers
80K /sbin/systemd-tmpfiles
60K /sbin/systemd-tty-ask-password-agent
13M total