Going from an inode to a file

We have discussed the fifteen entries in the block array and how they are used to specify data blocks which house a file. We have also seen that these sixty bytes may be repurposed for certain cases. If the inode stores a symbolic link and the target is less than sixty bytes long it will be stored in the block array. If the in-line data flag is set, the first sixty bytes of a file are in this space. We have also mentioned that extents use this space differently.

When parsing an inode for the list of blocks contained in a file (or directory) in the most generic situation (i.e. no extents), we read the first twelve entries in the block array and add any non-zero values to our list of blocks. If the thirteenth entry is non-zero, we read in the block with the address stored there which contains a list of up to 1024 blocks. Any non-zero entries from the data block are added to our list. This allows files as large as 4 megabytes + 48 kilobytes to be handled.

If there is an entry in the fourteenth block array item, we load the specified data block which contains a list of data blocks which in turn contain lists of data blocks. There are up to 1024 * 1024 entries in this doubly indirect level all of which point to 4 kilobyte blocks which allow files as large as 4 gigabytes + 4 megabytes + 48 kilobytes to be handled. If this is still not enough, the last entry is for a triply indirect block which permits files as large as 4 terabytes + 4 gigabytes + 4 megabytes + 48 kilobytes, at least in theory. If you have an ext4 filesystem, it is much more likely to use extents which are discussed in the next section.