On Wed, Jul 02, 2025 at 08:55:06AM +1000, Dave Chinner wrote: > .... these asserts will fire for single block XMBUF buffers because > PAGE_SIZE < mp->m_sb.sb_sectsize when the metadata sector size is > larger than PAGE_SIZE.... Oh right, we now support metadata sector sizes > PAGE_SIZE. > Also, I can see no obvious reason for getting rid of this > abstraction and none has been given. It doesn't make the code any > cleaner, and it introduces an incorrect implicit assumption in the > buffer cache code (i.e. that every buftarg has the same sector > size). Hence I don't think we actually should be "cleaning up" this > code like this... > > Now, if there's some other functional reason for doing this change > that hasn't been stated, let's talk about the change in that > context. Can you explain why this is necessary? Mostly because they way they were used is horribly confusing. The worst part is fixed up in patch 1, but I wanted to go one step futher. I guess I'll just drop this patch and we'll have to live with the two sector sizes.
