ls hangs for 90 seconds per each disconnected automount

[blsz]

ls hangs for 90 seconds per each disconnected automount. gnuls is not similarly blocked.

> time gnuls -al /media/myname/
total 8
drwxr-x---+ 4 root root 4096 Nov  7 12:16 .
drwxr-xr-x  4 root root 4096 Aug 23 16:52 ..
drwxr-xr-x  2 root root    0 Nov 21 23:12 corsair
drwxr-xr-x  2 root root    0 Nov 21 23:12 san1TB

real    0m0.001s
user    0m0.000s
sys     0m0.001s

> time ls -al /media/myname/
total 8
drwxr-x--- 4 root root 4096 Nov  7 12:16 .
drwxr-xr-x 4 root root 4096 Aug 23 16:52 ..
d????????? ? ?    ?       ?            ? corsair
d????????? ? ?    ?       ?            ? san1TB

real    3m0.327s
user    0m0.003s
sys     0m0.003s

This occurs regardless whether the automount is "loaded active waiting" or "loaded active running" as long as the mountpoint is not available. NFS mounts are affected as well.

In Ubuntu 25.10, the issue causes incoming SSH sessions to hang for 90s per each retry of each unavailable automount because the MOTD scripts make use of rust-coreutils stat and ls. https://bugs.launchpad.net/ubuntu/+source/base-files/+bug/2130843

I am opening this bug per the suggestion of @Alonely0 in #9072.

Some context:

> grep "/media/" /etc/fstab
/dev/disk/by-label/corsair /media/myname/corsair auto x-systemd.automount,nosuid,nodev,nofail,x-gvfs-show,x-gvfs-name=corsair,noauto,user  0 0
/dev/disk/by-label/san1TB  /media/myname/san1TB  auto x-systemd.automount,nosuid,nodev,nofail,x-gvfs-show,x-gvfs-name=san1TB,noauto,user   0 0

> mount | grep "/media/"
systemd-1 on /media/myname/corsair type autofs (rw,relatime,fd=65,pgrp=1,timeout=0,minproto=5,maxproto=5,direct,pipe_ino=13402)
systemd-1 on /media/myname/san1TB type autofs (rw,relatime,fd=67,pgrp=1,timeout=0,minproto=5,maxproto=5,direct,pipe_ino=13405)

The issue occurs in both states:

> systemctl | grep media-
  media-myname-corsair.automount         loaded active waiting   media-myname-corsair.automount
  media-myname-san1TB.automount          loaded active waiting   media-myname-san1TB.automount

and

> systemctl | grep media-
  media-myname-corsair.automount         loaded active   running         media-myname-corsair.automount
  media-myname-san1TB.automount          loaded active   running         media-myname-san1TB.automount
  media-myname-corsair.mount             loaded inactive dead      start /media/myname/corsair
  media-myname-san1TB.mount              loaded inactive dead      start /media/myname/san1TB

[ChrisDryden]

Did some investigating into this and there's not an easy fix because we currently use the std fs Metadata object that doesn't have a constructor and we would have to call fstat directly with this flag to have the same behavior but we would first need to refactor the library to use an Metadata object other than the std fs Metadata object