Why Do df and du Report Different Disk Usage on Linux? Explained

When operating Linux servers, you often need to check disk space; typically you use df -lh to view total and used space of each mounted filesystem, or du -sh [directory] to calculate the space occupied by all files in a specific directory. When using df and du , you may encounter inconsistent disk usage reports. For example, du may report that the total size of files under the root directory is 2 GB, while df shows that the mounted disk’s used space is 3 GB, 20 GB, or even more. This discrepancy can be caused by three reasons: 1. Reserved space To guard against emergencies, the Linux ext filesystem reserves a portion of disk space. The reserved amount can be viewed with tune2fs -l [dev_name] | grep "Reserved block count" (where dev_name is the device name). This reserved space is counted as used by df , causing inconsistency with du . You can adjust the reserved size using tune2fs -m [size] [dev_name] . 2. Phantom files du sums the sizes of files recorded by the filesystem, while df reads usage information from the superblock, reflecting which disk blocks are allocated. If a file is deleted but still held open by a process, du will not count it, but df will still show its blocks as used. Such phantom files and the processes holding them can be identified with lsof | grep deleted . The space is released when the process stops or is killed. 3. Uncounted files If the first two causes are ruled out yet the numbers still differ, another situation may be at play: when a directory containing data is mounted onto a new device, du does not see the pre‑existing data because it resides on the underlying filesystem, but the space is still occupied and counted by df . Consequently, df reports higher usage than du . To resolve this, carefully kill all processes using the directory with fuser -km [directory] , unmount it via umount [directory] , delete the hidden data, and then remount the device ( mount -t [type] [dev] [directory] ).