Does the I/O scheduler prioritize block requests before the requests go to the generic block layer?

willdye

In section 15, about I/O scheduling, one paragraph reads:

The I/O scheduler provides the interface between the generic block layer and low-level physical device drivers. Both the VM (Virtual Memory) and VFS (Virtual File System) layers submit I/O requests to block devices; it is the job of the I/O scheduling layer to prioritize and order these requests before they are given to the block devices.

Should the final sentence above instead end with "...before they are given to the low-level physical device drivers."? If not, then I'm confused as to where the I/O scheduler fits in the layering scheme.

coop

That is what the sentence means as written. I don't see any confusion

willdye

The sentence "The I/O scheduler provides the interface between the generic block layer and low-level physical device drivers" indicates this relationship:

Block --> I/O scheduler --> Low-level drivers

The sentence "the VM (Virtual Memory) and VFS (Virtual File System) layers submit I/O requests to block devices; it is the job of the I/O scheduling layer to prioritize and order these requests before they are given to the block devices" indicates this relationship:

VM/VFS --> I/O scheduler --> Block

So the first sentence puts the scheduler at a lower level than the block, but the second sentence puts the scheduler at a higher level than the block. Most layering schemes look like cake layers, one atop the other. In this case there's a cake layer called "block", and another layer called "scheduler" that is below and above "block". Strange cake.

coop

block layer != block device. Block device = Low-level-drivers. Block Layer = Generic kernel code.

Your confusion is one of semantics, not substance as best I can see.