On Fri, Sep 23, 2022 at 01:35:08PM -0400, Paul Moore wrote: > If a dentry is truly needed by EVM (a quick look indicates that it may > just be for the VFS getxattr API, but I haven't traced the full code > path), then I'm having a hard time reconciling that this isn't a > dentry operation. Yes, I get that the ACLs belong to the inode and > not the dentry, but then why do we need the dentry? It seems like the > interfaces are broken slightly, or at least a little odd ... <shrug> The dentry_operations are bit misnamed and should probably have been called dcache_operations, that is they are all about managing the dcache state and closely related operations. ACLs aren't like that.
