On Fri, Oct 01, 2021 at 02:35:32PM -0400, Gabriel Krisman Bertazi wrote: > I don't like the idea of having a flavor of a dentry comparison function > that doesn't invoke d_compare. In particular because d_compare might be > used for all sorts of things, and this fix is really specific to the > case-insensitive case. > > Would it be possible to fold this change into generic_ci_d_compare? If > we could flag the dentry as part of a parallel lookup under the relevant > condition, generic_ci_d_compare could simply return immediately in > such case. Not really - if anything, that's a property of d_alloc_parallel() call done by d_add_ci(), not that of any of the dentries involved...