How to use `findmnt --poll` in a script

Question

Raspberry Pi generally mounts two partitions at boot time; here's /etc/fstab:

proc            /proc           proc    defaults          0       0
PARTUUID=8a6020b5-01  /boot/firmware  vfat    defaults          0       2
PARTUUID=8a6020b5-02  /               ext4    defaults,noatime  0       1

I have an issue in that on several occasions, the /boot/firmware (vfat) partition has (apparently at random) become "un-mounted"; I have noted this in the output of lsblk --fs :

# 'boot/firmware' mounted correctly: 

$ lsblk --fs
NAME        FSTYPE FSVER LABEL  UUID                                 FSAVAIL FSUSE% MOUNTPOINTS 
nvme0n1
├─nvme0n1p1 vfat   FAT32 bootfs 146B-AD94                             444.8M    13% /boot/firmware
└─nvme0n1p2 ext4   1.0   rootfs ece25014-1cd7-4874-9bfc-f91e486bc000  364.9G     2% /

# 'boot/firmware' becomes "un-mounted": 

$ lsblk --fs 
NAME        FSTYPE FSVER LABEL  UUID                                 FSAVAIL FSUSE% MOUNTPOINTS 
nvme0n1
├─nvme0n1p1 vfat   FAT32 bootfs 146B-AD94                             444.8M    13% 
└─nvme0n1p2 ext4   1.0   rootfs ece25014-1cd7-4874-9bfc-f91e486bc000  364.9G     2% /  

After considering a few options to help "troubleshoot" this issue, I thought that using the --poll option with findmnt would be a good choice; i.e.:

$ nohup /usr/bin/findmnt -n --poll=umount,mount --target /boot/firmware & 

I've run this command, and tested it using umount and mount to trigger findmnt --poll:

$ nohup /usr/bin/findmnt --poll=umount,mount --target /boot/firmware &
[1] 10904

# wait a while, then

$ sudo umount /boot/firmware 
$ sudo mount -a

# check nohup.out: 

$ cat nohup.out
umount     /boot/firmware /dev/nvme0n1p1 vfat   rw,relatime,fmask=0022,dmask=0022,codepage=437,iocharset=ascii,shortname=mixed,errors=remount-ro
mount      /boot/firmware /dev/nvme0n1p1 vfat   rw,relatime,fmask=0022,dmask=0022,codepage=437,iocharset=ascii,shortname=mixed,errors=remount-ro 

What I need to do is trigger a routine when when the umount ... output is produced. So far I've been unable to do that, and that's where I need some help. I need a script that starts the findmnt --poll command, and then waits for an output beginning with umount. When that output arrives, I'd like to run dmesg (?), capture that output to a logfile, maybe some other diagnostic(s), and finally issue a sudo mount -a.

After findmnt --poll outputs another line with mount, I think the script should return to a "waiting" state to capture the next umount event.

I've read a bit on the wait command, but wonder if it's the correct approach because:

The wait command makes a shell script or terminal session wait for background processes to finish.

In this case, findmnt --poll doesn't finish - it continues to run, polling for another reportable state change. Any suggestions on how to approach this?

Answer 1

Thanks to @muru's suggestion, I've got an answer:

#!/usr/bin/env bash

POLLMNT_LOG='/home/pi/pollmnt.log'

/usr/bin/findmnt -n --poll=umount,mount --target /boot/firmware |
while read firstword otherwords; do
    case "$firstword" in
        umount)
            echo -e "\n\n ===> case: umount" >> $POLLMNT_LOG
            dmesg --ctime --human >> $POLLMNT_LOG
            mount -a
            ;;
        mount)
            echo -e "\n\n ===> case: mount" >> $POLLMNT_LOG
            dmesg --ctime --human >> $POLLMNT_LOG
            ;;
        *)
            echo -e "\n\n ===> case: * (UNEXPECTED)" >> $POLLMNTLOG
            ;;
    esac
done