/run or /tmp

david5MX53G

What's the difference between /run and /tmp?

"The purpose of /run is to store transient files: those that contain runtime information, which may need to be written early in system startup, and which do not need to be preserved when rebooting."

"/tmp
This directory is used to store temporary files, and can be accessed by any user or application."

One is for "transient files" (existing, happening, or staying somewhere for a short period of time only) and one is for "temporary files" (existing, done, or used for only a limited period of time).

Why don't all the distros simply use /tmp?

luisviveropena

Hi @david5MX53G , that's an interesting question!

/tmp is much older than /run/ . The first is for unprivileged users, they can't write on /run. /run is used for services.
There is a pretty interesting discussion here:

https://unix.stackexchange.com/questions/316161/whats-the-difference-between-tmp-and-run

Regards,
Luis.

coop

The stackexchange discussion is useful but there is some attitude bias in some of the posts. For day to day use when you don't really care about security, or on a single user system (which almost students for this course are!) /tmp is just fine and easier to use than trying to put files in /run/user/1000 (substitute your ID if it is not 1000). However, keep in mind two other facts:

1) on Ubuntu /tmp is erased every time you boot while most other distros (including anything in the red hat family and debian) it is preserved. However, anything not touched might be erased after (10) days, at least in the Red Hat family.

2) Some distros (Fedora by default) put /tmp in a tmpfs filesystem (ram disk) so if you make big files they use up your memory -- /run does the same thing. So don't put some very large file in /run or on Fedora in /tmp unless you change that boot behaviour (how to avoid it is in the course somewhere (sudo systemctl mask tmp.mount , and then reboot)

Personally, I never put anything directly in /run although the system will do it for me sometimes. But that is just me.

david5MX53G

Although the following rationale is not given in the FHS, /run was introduced to overcome the problem that /var was mounted too late such that dirty tricks were needed to make /var/run available early enough.

That makes me wonder: why is /var mounted "too late" and what does "too late" mean? Can't /var just be mounted before running whatever stuff needs it? ...or is /var a world-writable dir like /tmp?

coop

AFAIK, the only world-writable part of /var is /var/tmp. I'm not personally sure why applications use it; on my RHEL 8 system it seems to be full of useless files from "flatpak" which RHEL8 forced me to install and which I have never used (lets not go there.) but also some stuff from dnf. I'm sure I can obliterate all these files (they are "tmp") and I see they are up to two months old at the moment so the system has not yet rinsed them for me. I have seen distribuctions where you have ln -s /tmp /var/tmp anyway.

This discussion of "too early" etc really is historical. Once upon a time it was very common to have a system boot only using /bin /sbin etc and mounting /usr (and possibly /var, /home etc) later, perhaps through a network. I grew up on Solaris type systems where /home was on NFS on a server for example. Today these are not common systems and all RHEL-based systems now have /bin->/usr/bin and the same for sbin lib and lib64. As is noted in the course the idea that /usr is mounted after boot is gone -- the root filesystem contains both /usr and /var and /home. I just checked and the same thing is true on Ubuntu 20.04, so I guess all systemd-based systems now have this.

You notice you say "was" mounted not "is" mounted. So today this is all "meh"