Boot script to check for dirty bit/corruption on a USB stick or SD card

Question

I posted this question on the Raspberry Pi forum but got no activity. If it is a dumb question or something else, please say so...

While not often, I do experience power outages which stop one of my Pis and then usually results in the USB stick having the "dirty bit" set.

I know how to fix this manually:

sudo dmesg |grep sda1 ==> this finds references to sda1 in the message log.

If I see that it has reported a dirty bit for sda1 then I do the following:

1. sudo umount /boot

2. sudo fsck -V /dev/sda1 ==> answer questions based on the following output:

   fsck.fat 3.0.24 (2013-11-23)
   0x25: Dirty bit is set. Fs was not properly unmounted and some data may be corrupt.
   1) Remove dirty bit
   2) No action
   

   I reply this: ? 1
   More output:

   Starting check/repair pass.
   Starting verification pass.
   Leaving filesystem unchanged.
   /dev/sda1: nn files, mmm/ooo clusters
   

3. sudo fsck -a /dev/sda1

4. mount /boot

However, I would like to automate this in a script which could be run manually or during boot up. Can anyone assist or point me to a good reference? Obviously, I may also want to do this for the SD card OS partition.

Answer 1

You don't really need a script to detect a dirty bit. Just do the check every time: fsck is smart enough not to touch a file system if there's nothing to fix.

Native Linux file systems have a built-in option telling the system how often they should be checked. If you want your root FS to be checked on every tenth reboot, run

tune2fs -c 10 /dev/mmcblk0p2

Another option to consider is journaling, which will eliminate the need for frequent fsck checks, at the expense of increased wear of your SD card.

Answer 2

Ran into the same problem, and your question contained the answer I needed!

Since I now need the same script as you, I thought of paying you back :-) (hopefully you are still interest after over a year later)

I'm not a pro, so this might be highly unefficient, but it seems to work for me:

1) Save your first command's output to a temp file

sudo dmesg |grep sda1 >> /home/pi/temp_file

2) Check if there's the "dirty bit set" message in the temp file

if grep "Filesystem has been set read-only" /home/pi/temp_file

3) If it's the case, unmount (same command), run fsck with the automatic fixing without prompt option (-a)

sudo umount /dev/sda1
sudo fsck.vfat -a /dev/sda1
sudo reboot

4) You should delete the temp file before rebooting and before exiting the script. You can also add logging for good measure.

Full example:

#!/bin/bash

date >> /home/pi/test_sh.log
sudo dmesg |grep sda1 >> /home/pi/temp_file
if grep "Filesystem has been set read-only" /home/pi/temp_file
then
echo "Dirty bit set!" >> /home/pi/test_sh.log
rm /home/pi/temp_file
sudo umount /dev/sda1
sudo fsck.vfat -a /dev/sda1
sudo reboot
else
echo "Everything ok" >> /home/pi/test_sh.log
rm /home/pi/temp_file
fi
exit

(Note: I have this running once a day and the log purged once a week. It also runs on boot so just rebooting triggers the script. My SO has the habit of rebooting the RPi when it acts up :-P)