33

I tried to fix my hard drive using Ubuntu but I encountered an error message and didn't know how to proceed. I need to fix my bad sectors

fsck /dev/sdb  
fsck from util-linux 2.20.1  
e2fsck 1.42.5 (29-Jul-2012)  
fsck.ext2: Permission denied while trying to open /dev/sdb  
You must have r/w access to the filesystem or be root  
1
  • 2
    The text you've pasted doesn't seem related to your question - can you clarify? Jan 14, 2013 at 2:39

4 Answers 4

31

A bad sector on a drive is a sign of permanent damage to the drive. Unless you have reason to believe that your drive marked these sectors as bad incorrectly, you cannot "fix" them.

It means that a part of your drive is damaged to the extent that it can no longer reliably be read from and/or written to.

Your system can continue to use the drive by marking that sector as unusable, but if you have enough bad sectors, or a SMART tool triggers a warning level, you might consider a drive replacement, as bad sectors can be a sign that more sectors, or the whole drive, might fail soon.

While there may be ways to force the drive to un-mark a sector as bad, allowing you to use it again, this is likely not a good idea. The sector may stay good, but it will just as likely become bad again. Some data may be lost or corrupted depending on how it fails.

Now, as for the error message you've pasted in your question (as of my writing this), that error has nothing to do with bad sectors. It means that you don't have access to the drive. Being sudo can give you access, so:

sudo fsck /dev/sdb

However, this is still probably not what you want, because /dev/sdb refers to the entire drive, whereas fsck is designed to work on filesystems, which are usually (but not always, and you may have an exception here) placed in partitions. If the above didn't work, you may instead have wanted to do this to the 1st partition on that drive:

sudo fsck /dev/sdb1

You can get a list of partitions per drive with:

sudo fdisk -l
6
  • 9
    Its probably just a bad block. It can't read the data, is all. After remapping the block the drive is likely good for another three years. So goes my experience.
    – Zan Lynx
    Jan 27, 2015 at 20:53
  • 2
    A HDD with a bad sector is much, much more likely to fail at any moment than one which has never had a bad sector. How many drives is your experience based on? I would still call it a bigger gamble. Jan 27, 2015 at 23:45
  • 1
    More likely? Sure. But it isn't immediately doomed. So much depends on why that block went bad. Like I said, I've had drives that slowly remap a few bad blocks over their lifetime and just keep working.
    – Zan Lynx
    Jan 28, 2015 at 0:00
  • 1
    it isn't immediately doomed - you have no way of knowing or guaranteeing that. In some cases it might be. So much depends on why that block went bad - it would be impossible to reliably find out the why, without opening it up and effectively destroying it. As I said it's always an increased risk. If you are absolutely fine with such a risk I have no problem with that. Most of the time, maybe even >90% of the time, the drive won't quickly fail after a bad sector is found and in some cases it might even last for years after. Keep good backups and any future failure won't hurt as much. Jan 28, 2015 at 0:50
  • 1
    Says "disk is mounted" and does nothing.
    – Dims
    Mar 24, 2016 at 19:59
28

In short: Boot a rescue system and use badblocks:

badblocks -svn /dev/sda

to have the hard disk controller replace bad blocks by spare blocks.

I have answered the same question in detail here.

3
  • 6
    You probably want to use this indirectly with fsck's -c option. AFAIK using badblocks standalone will never change the behaviour of anything, it will only report on where the bad blocks are. Of course, the drive firmware might transparently modify its behaviour when put under the usage patterns characteristic of badblocks, but surely this would be drive dependent?
    – amoe
    Nov 11, 2015 at 14:13
  • 1
    As a new user you first need to gain "reputation" before being allowed to do certain activities on this site. So it was not possible to duplicate this. Still the question deserved an answer. As bad blocks are an issue of the physical disk (check out SMART values, the filesystem is not directly affected. It might suffer corruption due to data loss. So doing a fsck is a good idea as well. As original question was about how to "repair" bad blocks this was the answer.
    – user228505
    Feb 1, 2016 at 2:24
  • 2
    @amoe the documentation supposes the behaviour will change, here: "Run sudo badblocks -n on a device name to run it by itself and report badblocks right in your terminal. You can use the -w option to use a write-mode test, but don’t use the -n and -w options together since they’re mutually exclusive. You should under no circumstance ever use the -w option on a volume that has data, since it will erase everything clean. Use the slower -n option since it will preserve your data in this case. The -w option is fine for volumes that you don’t mind erasing."
    – DiaJos
    Jul 17, 2019 at 19:56
19

In the olden days you used to have to take a note of the badblocks that were written on the drive and enter those in the defect list before formatting the drive, because hard disk surfaces were never "perfect", manufacturer's got wise to this because people buying disks would look at the defect table printed on the disk and buy the ones with the least amount of defects...

Now scroll forward 20 or so years and hard disk manufacturers hide the fact a brand new disk has bad blocks with the firmware, when you buy a brand new disk it will have in all probability bad blocks already, the firmware will detect newly grown badblocks and maps them out from a set of spare cylinders it has, but this only happens when a write operation occurs on that sector and the ECC algorithm detects bit failure, only then will it map the block out. So getting back to the point you can force a drive to map out the badblocks by simply using DD, ie/ [edit: following example was edited to prevent accidental drive destruction: removed wildcard character, replaced with X].

WARNING: DO NOT BLINDLY USE THE dd COMMAND BELOW IF YOU DO NOT KNOW WHAT IT DOES.

dd if=/dev/zero of=/dev/sdX bs=1M 

obviously alter the of=target to reflect the drive you want to zero this is the quickest way to re-map a drives defects, its also the quickest way to totally trash your Linux setup by getting the target drive wrong, so check, double check then check again, before you press the enter key. You can increase the size of the blocksize in the dd command to optimize read/writes and perhaps make things go quicker, but there are diminishing returns after a point. I find anything between 1M and 8M works best for me

You can get DD to just write one sector, the bad one... to get it remapped, so you don't have to backup your drive, but that's a whole different kettle of fish and Russian roulette if you don't know exactly what you're doing .....

This command WILL destroy everything on your hard drive including any partitions. But it will force the drive to map out any bad sectors it may have.

It is perfectly ok for a disk to have bad sectors as long as they aren't on the boot sector, if they are then the drive is useless, if you notice a few months later that your drive has developed more bad sectors then it's time to start shopping for a replacement.

I've revived hundreds of drives like this and they've lived on for many years afterwards, a bad sector isn't necessarily the death knell for the drive.

2
  • I've used this method before but never knew why it worked. Out of a handful most drives lasted for years after shallow wiping the drive as prescribed.
    – MER
    Oct 30, 2016 at 4:58
  • 4
    Ack!! For the uninitiated, PLEASE DO NOT COPY AND PASTE THAT LINE. Your shell will expand the wildcard '?' with the actual device names in /dev. If you should accidentally hit enter before replacing the ?, you could possibly wipe the wrong drive. If you have more than 1 device, dd might throw an error, but I am unwilling to test that on my machine. Jun 25, 2018 at 17:53
1

After "man badblocks" and "man e2fsck" I found:

You probably do not want to run badblocks as a standalone but with e2fsck. To run e2fsck with badblocks add the -c option. "This option causes e2fsck to use badblocks program to do a read-only scan of the device in order to find any bad blocks. If any bad blocks are found, they are added to the bad block inode to prevent them from being allocated to a file or directory."

You can also add -k to add to the bad block list and not create a new one.

sudo e2fsck -ck /dev/sda1

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .