On Thu, Jul 17, 2025 at 09:51:23AM +0200, Christoph Hellwig wrote: > On Thu, Jul 17, 2025 at 09:48:49AM +0200, Christian Brauner wrote: > > On Wed, Jul 16, 2025 at 03:02:00PM +0200, Christoph Hellwig wrote: > > > On Tue, Jul 15, 2025 at 05:09:03PM +0100, Matthew Wilcox wrote: > > > > will be harder, we have to get to 604 bytes. Although for my system if > > > > we could get xfs_inode down from 1024 bytes to 992, that'd save me much > > > > more memory ;-) > > > > > > There's some relatively low hanging fruit there. > > > > > > One would be to make the VFS inode i_ino a u64 finally so that XFS > > > and other modern files systems an stop having their own duplicate of > > > > That's already on my TODO since we discussed this with Jeff last year. > > Cool! > > Btw, I remember anothing I've been wanting to look at, which is > killing the u/g/p quota pointers. If we used a rhashtable with > proper sizing for them, doing a hash lookup instead of the caching > should be efficient enough to be noise compared to the actual quota > operations. That would free three pointers per inode, or in case > of XFS six without the optimization in this thread. That would be really nice. :) --D
