On Fri, Feb 07, 2025 at 10:12:07AM +0100, Daniel Gomez wrote: > On Thu, Feb 06, 2025 at 02:50:28PM +0100, Luis Chamberlain wrote: > > On Thu, Feb 06, 2025 at 02:27:16PM -0800, Darrick J. Wong wrote: > > > Now you've decreased the default blocksize to 512 on sda, and md0 gets > > > an impossible 512k blocksize. Also, disrupting the default 4k blocksize > > > will introduce portability problems with distros that aren't yet > > > shipping 6.12. > > > > Our default should be 4k, and to address the later we should sanity > > check and user an upper limit of what XFS supports, 64k. > > To clarify, the patch addresses the cases described above. It sets mkfs.xfs's > default block size to the device’s reported minimum I/O size (via stx_blksize), > clamping the value between 4k and 64k: if the reported size is less than 4k, 4k > is used, and if it exceeds 64k (XFS_MAX_BLOCKSIZE), 64k is applied. Ah, neat. Then the only other concern expresed by Darrick is using this for distros not yet shipping v6.12. Let me address that next in reply to another part of this thread. Luis
