On Tue, 08 Oct 2013 13:25:29 -0700, Guangyu Sun said:

As far as I know sector size comes with the storage device and cannot be changed, while block size is set by file system.

On some high-end RAID-based storage devices, the effective sector size is basically the raid stripe size and selectable (for the DDN SFA10K you get a choice of 128K, 256K, or 1M chunks on a per-LUN basis, and writing anything smaller requires a read-modify-write cycle). But this is the sort of stuff that the vast majority of Linux users won't ever encounter, because you'll probably never have a petabyte-sized filesystem. (Protip: mmfsck of a 700T GPFS filesystem with 220M files takes about 32 hours. Don't ask how I found this out. :)