On Wed, Feb 19, 2025 at 10:13:24PM -0800, Christoph Hellwig wrote: > On Wed, Feb 19, 2025 at 10:44:09AM +0800, Ming Lei wrote: > > PAGE_SIZE is applied in validating block device queue limits, this way is > > very fragile and is wrong: > > It's neither very fragily nor wrong. If you want to change it to suit > your needs that might or might not be ok but this language isn't. max segment size is fixed value since it is read from hardware, now kernel validates it by variable PAGE_SIZE, not fragile? Well, it isn't my need only, there are lots of device which max segment size is < 64K, they work just fine in 4K page size kernel, however, they become unusable in 64K page size kernel. > > > - queue limits are read from hardware, which is often one readonly hardware > > property > > queues limits aren't read from hardware per definition. Very often they > are software limits. Fine, I will update the commit log to just mention max segment size limit. Thanks, Ming
