I'm currently trying to mount any usb drive connected to my computer automaticaly. My goal is to mount usb devices either with the label if they have one or with the uuid if they don't.
To do it I write a udev rule in /etc/udev/rules.d/10-usb-detect.rules :
ACTION=="add", KERNEL=="sd?[0-9]", SUBSYSTEM=="block", RUN+="/usr/local/bin/add.sh"
The script is called everytime an add event append on the block subsystem.
The udev rule works fine however when I try to mount the file system from the script it doesn't work. What is strange is that the mount command from the script always return $?=0, so logically the file system should be mounted but it's not.
Here is my script :
#!/bin/bash
LOG_FILE=<path_to_file>
echo "New usb device detected at $DEVNAME" >> $LOG_FILE
echo "mount $DEVNAME /media/usb/test" >> $LOG_FILE
mount $DEVNAME /media/usb/test &>> $LOG_FILE
ret=$?
echo "$ret" >> $LOG_FILE
if [ $ret == "0" ]; then
echo "$DEVNAME mounted at /media/usb/test" >> $LOG_FILE
else
echo "Failed to mount $DEVNAME at /media/usb/test" >> $LOG_FILE
fi
echo "" >> $LOG_FILE
I tried with /media/usb/test not existing and I have the expected error from the mount command. But when the folder exists he mount command returns 0 even if the filesystem is not mounted.
Here is an output of the log file :
New usb device detected at /dev/sdc1
mount /dev/sdc1 /media/usb/test
mount: mount point /media/usb/test does not exist
32
Failed to mount /dev/sdc1 at /media/usb/test
New usb device detected at /dev/sdc1
mount /dev/sdc1 /media/usb/test
0
/dev/sdc1 mounted at /media/usb/test
/dev/sdc1 is not mounted even if mount returns 0.
I precise that when I mount the file system from command line, there is absolutely no problem and the file system is mounted as it should.
Does someone has a clue about how I could debug this?
I think the problem is because the script is called from udev because if I call it from command line it also works.
I think I identified the problem.
Apparently, udev uses a specific namespace, indeed I can see the mount point by printing the content of /proc/<daemon_pid>/mountinfo where is the pid of the systemd-udevd service.
$ cat /proc/240/mountinfo
[...]
228 43 8:17 / /media/usb/test rw,relatime - vfat /dev/sdb1 rw,fmask=0022,dmask=0022,codepage=437,iocharset=ascii,shortname=mixed,utf8,errors=remount-ro
$ df
Sys. de fichiers blocs de 1K Utilisé Disponible Uti% Monté sur
udev 1975740 0 1975740 0% /dev
tmpfs 397376 5916 391460 2% /run
/dev/sda2 75733088 69473400 2389532 97% /
tmpfs 1986868 111860 1875008 6% /dev/shm
tmpfs 5120 4 5116 1% /run/lock
tmpfs 1986868 0 1986868 0% /sys/fs/cgroup
tmpfs 397372 28 397344 1% /run/user/112
tmpfs 397372 24 397348 1% /run/user/1001
So the solution should be to force udev to execute the script in root userspace. I tried the solution I found here https://unix.stackexchange.com/questions/330094/udev-rule-to-mount-disk-does-not-work
However, my system doesn't have a /usr/lib/systemd/system/systemd-udevd.service file. I created a file /etc/systemd/system/systemd-udevd.service with content
MountFlags=shared
But with this solution, my system is not able to boot anymore.
Does someone know how I could either execute the script in root userspace or share the mount point with users?
PS : I precise I'm running a 64 bits Debian 9
Solved edit : Finally the file was located at /lib/systemd/system/systemd-udevd.service. I duplicated it in /etc/systemd/system/systemd-udevd.service
and changed MountFlags=slave to MountFlags=shared and now it works perfectly :)