On Mon, Mar 16, 2020 at 05:51:54PM +0100, Christoph Hellwig wrote: > > I've also got a patch deleting these confusing HMM_PFN_ERRORs. They > > are not applied consistently and no caller would scan the output for > > ERROR on some failures. For instance the above doesn't set it on EPERM. > > > > So this can just be 'return -EFAULT' instead of 'goto out_error' > > > > I can fold that in if you agree: > > If you want to fix this up and the other ->pfns initialization feel > free. Otherwise I'd be happy to resend. While we are at it: hmm_vma_walk_hole_ in the subject should be hmm_vma_fault. I renamed the function on the last rebase as the old name was so horrible.