inconsistent filesystem traversing

[leszekeljasz]

Hello!

Thank you for this lovely tool. Just spotted this behaviour that does not seem to be desired. Here's my situation, a Linux server with two mounts: / and /var/www/virtualmin:

root@web01:/# df -h | grep -vE 'snap|run|udev|tmpfs'
Filesystem      Size  Used Avail Use% Mounted on
/dev/nvme0n1p1   39G   16G   24G  40% /
/dev/nvme1n1     30G   25G  4.7G  85% /var/www/virtualmin

When I use vmtouch against /, it indicates total pages ~ 16G which aligns with filesystem usage on the / mount - so looks like vmtouch skips the mount in /var/www/virtualmin:

root@web01:/# vmtouch / 2>&1 | grep -v WARNING
           Files: 544678
     Directories: 47977
  Resident Pages: 209724/4307663  819M/16G  4.87%
         Elapsed: 47.997 seconds

But when I execute it against all files in / (skipping some), it indicates total pages as 39GB which more or less aligns with total filesystem usage on both / and /var/www/virtualmin, so in this variant it includes the mount:

root@web01:/# vmtouch $(ls -d /* | grep -vE 'dev|sys|proc|snap|run') 2>&1 | grep -v WARNING
           Files: 819215
     Directories: 103000
  Resident Pages: 305073/10413686  1G/39G  2.93%
         Elapsed: 45.924 seconds

Is this expected? My linux version: 5.4.0-1103-aws.