Blog format rewritten, write blog too slow low efficiency using the previous method, although you can review it again intact, but the efficiency is really low.
The system will query only block device, does not query file (that is, the system will not query the device's file, only direct access to block devices, as a device block device I guess that part of occupied blocks)
When you use -sh ll, it is at the upper left total (Chinese is: the total amount) that is in fact the actual number of blocks of all data in the directory consumed * block size value
Here anaconda-ks.cfg and initial-setup-ks.cfg these two documents is a size of a 1.8k 2.0k but with the size of 8k (because a block is 4k) which will result in a waste of disk .
So every time you create a file, that they will waste a lot of capacity, such as inode table, super block and other (mess hehe) data will cause a lot of wasted capacity