A couple of ring buffer questions

[bateyejoe]

Two questions on ring buffers.

1. In io_uring_register_buf_ring(), the address of the buffer ring needs to be page-aligned. This means I can have only one ring per system page size of memory. If I want to create many small rings, this seems to waste a lot of memory and mapped address space. Is there a reason I can't allocate a page-aligned page of memory and create multiple rings within it rather than just one at the beginning of the page?

2. Prior to the buffer rings, "provided buffers" could be registered with io_uring_register_buffers() and similar methods to pre-map the buffers into the kernel. With the buffer ring, where does the mapping/unmapping occur on buffers placed into the buffer ring and completed by the multishot operations? Is there any way to pre-map these buffers or even any advantage to doing so?