"inode_size (128) * inodes_count (0) too big for a filesystem with 0 blocks" while creating a file system

Question

My partition table looks like this:

 Device Boot      Start         End      Blocks   Id  System
/dev/sda1   *        2048    32505855    16251904   83  Linux
/dev/sda2        32505856    33554431      524288   83  Linux

When I went to lay down a filesystem on sda2, it threw this error:

sudo mkfs -t ext4 /dev/sda2
mke2fs 1.42.9 (4-Feb-2014)
mkfs.ext4: inode_size (128) * inodes_count (0) too big for a
filesystem with 0 blocks, specify higher inode_ratio (-i) or lower inode count (-N).

I have tried both with an extended partition and a primary partition and get the same error. I have Ubuntu 14.04TLS. What to do?

Solution

1: it doesn't have to do anything with primary/extended/logical partitions.

2: I think you wanted to say "logical" partition instead of "extended".

3: mkfs thinks your partition size if 0 bytes. It was very surely, because the kernel wasn't able to update the partition table after a repartitioning. After you edited the partition table, didn't you get some warnings about that a reboot is needed?

On Linux, there is two different partition table: there is one on the zeroth block of the hard disk. And there is one in the kernel memory. You can read the first with an fdisk -l /dev/sda command. And the second can you read with a cat /proc/partitions command. These two need to be in sync, but it is not always possible. For example, you can't change the limits of a currently used partition. In this case, the kernel partition table won't changed.

You can let the kernel re-read the disk partition table with the command blockdev --rereadpt /dev/sda. Most partitioning tools execute this command after they wrote out your newly changed partition table to the disk.

The problem is that only newer linux kernels are capable to re-read a partition table of a used hard disk. From this viewpoint, a hard disk is considered as "used" if there is a simple partition which is used on it, either by a tool, or a mount point or it is an active swap partition.

And even these newer kernels aren't able to change the limits of a partition currently being used.

I think, your root system is on /dev/sda, thus you need to do a reboot after you did a repartitioning.

Comments

[Admin]

You reached 15 point of reputation, from now you can upvote questions (next to accepting them), if you are satisfied with them.

[user53029]

Yep, you're right. I haven't rebooted. So as you can see I wiped out my swap partition and commented it out in fstab, so do you think there will be an issue with the reboot? I have been told you really don't need a swap partition.

[user53029]

also, I ran the blockdev command and it failed as well so cat /proc/partitions still reflect the old kernel mapping. The error I got after running the blockdev command was a resource busy error - like I needed to reboot for that as well. Kernel is 3-13-0-29

[peterh]

@user53029 Sorry for the late react - ein blockdev --rereadpt /dev/sdX re-reads the partition table, if it is possible (and gives an error if it is not).