[Discuss] Linux command-line mystery: why does "cd" hang?

[dbarrett at blazemonger.com (Daniel Barrett)]

Thanks to Gregory Galperin, I was able to zero in on the culprit:

  /etc/profile.d/80-systemd-osc-context.sh

which is a symlink to:

  /usr/lib/systemd/profile.d/80-systemd-osc-context.sh

which causes nasty magic to execute with every bash prompt. ("bash
--norc" relieved the symptom.) The solution, with credit to Gemini,
was:

# rm /etc/profile.d/80-systemd-osc-context.sh
# ln -s /dev/null /etc/tmpfiles.d/20-systemd-osc-context.conf

TBH, I don't deeply understand the root cause or the fix, or why this
particular directory triggered the issue but not other directories.
The fix worked though.

Dan

On September 8, 2026, Daniel Barrett wrote:
>If you like Linux puzzles, I've encountered a command-line mystery
>like nothing I've ever seen before.
>
>tl;dr: I have a directory with mode 0755. Anytime I cd into it, my
>shell hangs. You might think, "Oh, that's a hung mount point," but no.
>It's just a subdirectory within the root filesystem, and the root user
>can cd into it with no problems. You might think, "Oh, maybe your shell
>prompt PS1 has some funky stuff that's not returning," but no, my prompt
>is just a static dollar sign. Any other guesses?
>
>Full details:
>
>I'm logged in via ssh on a Linode server running Ubuntu 26.04. On that
>server, I have a WordPress-related directory in my web tree,
>/var/www/wordpress/wp-content/uploads:
>
>  $ cd /var/www/wordpress/wp-content/uploads
>  <succeeds>
>
> It has a subdirectory, prime-mover-import-files. It appears to be an 
> ordinary directory:
>
>  $ ls -ld prime-mover-import-files
>  drwxr-xr-x 2 www-data www-data 12288 Sep  8 20:53 prime-mover-import-files
>
>  $ df prime-mover-import-files/
>  Filesystem     1K-blocks    Used Available Use% Mounted on
>  /dev/sda        25186308 5130400  18755860  22% /
>
>I don't see any unusual attributes:
>
>  $ stat prime-mover-import-files
>    File: prime-mover-import-files
>    size: 12288           Blocks: 24         IO Block: 4096   directory
>  Device: 8,0     Inode: 270434      Links: 2
>  Access: (0755/drwxr-xr-x)  Uid: (   33/www-data)   Gid: (   33/www-data)
>  Access: 2026-09-08 21:05:43.038735433 +0000
>  Modify: 2026-09-08 21:05:43.038735433 +0000
>  Change: 2026-09-08 21:05:43.038735433 +0000
>   Birth: 2026-09-08 19:37:29.024640959 +0000
>
>  $ stat -f prime-mover-import-files
>    File: "prime-mover-import-files"
>      ID: 738537e1353f0990 Namelen: 255     Type: ext2/ext3
>  Block size: 4096       Fundamental block size: 4096
>  Blocks: Total: 6296577    Free: 5016423    Available: 4691411
>  Inodes: Total: 1593088    Free: 1428275
>
>  $ lsattr -d prime-mover-import-files
>  -----------I--e------- prime-mover-import-files
>
>At the current moment, an hours-long WordPress import operation is in
>progress (over https) involving the prime-mover-import-files
>subdirectory. During that time, when I cd into the directory, my shell
>hangs.
>
>  $ cd prime-mover-import-files
>  <shell hangs>
>
>It hangs so badly that ^C has no effect, and I can't even terminate my
>SSH connection with the ssh escape sequence "~." I can only kill the
>enclosing X window. But the root user can cd into the directory anytime.
>
>  # cd prime-mover-import-files
>  # ls
>  <files are listed>
>
>I thought, maybe the system was so busy it couldn't serve my cd
>command, so I waited about 5 minutes. Still hung. Also, all other
>commands in other windows work fine.
>
>I checked the process state of the hung bash shell from another shell:
>
>  # ps -ut4
>  USER     PID %CPU %MEM    VSZ   RSS TTY      STAT START   TIME COMMAND
>  me     63169  0.0  0.5   8968  5844 pts/4    Ss+  21:18   0:00 -bash
>
>I killed that shell as root (kill -9 63169) but my session was still hung.
>
>ChatGPT suggested that the problem has to do with my prompt, PS1, not
>returning. My prompt for both users (root and non-root) was:
>
>  # echo $PS1
>  \[\e]0;\u@\h: \w\a\]${debian_chroot:+($debian_chroot)}\u@\h:\w\$
>
>However, I changed my non-root prompt to something static:
>
>  PS1='$ '
>
>then logged out, logged back in, and the problem still occurred. Plus,
>the root user has the fancy PS1 value and it encounters no problems
>cd-ing into the subdirectory in question.
>
>ChatGPT also suggested a hung mount point, but it's just a normal
>subdirectory, and no other commands hang, and root can cd into it.
>
>The problem appears to vanish when the hours-long WordPress import is
>finished. I can only guess that WordPress has temporarily restricted
>the directory in some weird way that doesn't affect root??
>
>I'm stumped. Any ideas what I should check next?
>
>Thanks very much,
>Dan
>
>--
>Dan Barrett
>dbarrett at blazemonger.com