What's the usage of blk_queue_segment_boundary()

Question

The book ldd says for the function blk_queue_segment_boundary() like this:

Some devices cannot handle requests that cross a particular size memory bound- ary; if your device is one of those, use this function to tell the kernel about that boundary. For example, if your device has trouble with requests that cross a 4- MB boundary, pass in a mask of 0x3fffff. The default mask is 0xffffffff.

I don't quite understand what the boundary means here, for example, I have a virtual block device, which are made of indeed 4MB files, so I want a request not exceed 4MB boundary,

unsigned long sector = blk_rq_pos(req);
unsigned long offset = sector << 9;
unsigned long nbytes = blk_rq_bytes(req);

int file_offset = offset % (1 << 22);

What I want is that (file_offset + nbytes) not greater than 4M, but indeed sometimes it exceeds 4M, so, is there any misunderstanding of blk_queue_segment_boundary() ?

Answer 1

Some controllers (particulary IDE) can not handle DMA requests that cross memory regions at 4MB. Think of it as segment:index addressing where index can not be larger that the set boundary.

There is also a blk_queue_max_segment_size. Both are used to construct correct requests to the device - requests get reordered and merged.

There are other uses. For example, from xen-blkfront.c:

/* Each segment in a request is up to an aligned page in size. */
blk_queue_segment_boundary(rq, PAGE_SIZE - 1);
blk_queue_max_segment_size(rq, PAGE_SIZE);

Requests are limited to PAGE_SIZE for better performance.