On 2021/08/28 1:43, Christoph Hellwig wrote: > On Fri, Aug 27, 2021 at 02:28:58PM +0000, Tim Walker wrote: >> There is nothing in the spec that requires the ranges to be contiguous >> or non-overlapping. > > Yikes, that is a pretty stupid standard. Almost as bad as allowing > non-uniform sized non-power of two sized zones :) > >> It's easy to imagine a HDD architecture that allows multiple heads to access the same sectors on the disk. It's also easy to imagine a workload scenario where parallel access to the same disk could be useful. (Think of a typical storage design that sequentially writes new user data gradually filling the disk, while simultaneously supporting random user reads over the written data.) > > But for those drivers you do not actually need this scheme at all. Agree. > Storage devices that support higher concurrency are bog standard with > SSDs and if you want to go back storage arrays. The only interesting > case is when these ranges are separate so that the access can be carved > up based on the boundary. Now I don't want to give people ideas with > overlapping but not identical, which would be just horrible. Agree too. And looking at my patch again, the function disk_check_iaranges() in patch 1 only checks that the overall sector range of all access ranges is form 0 to capacity - 1, but it does not check for holes nor overlap. I need to change that and ignore any disk that reports overlapping ranges or ranges with holes in the LBA space. Holes would be horrible and if we have overlap, then the drive can optimize by itself. Will resend a V7 with corrections for that. -- Damien Le Moal Western Digital Research
