On Fri, Oct 27, 2023 at 9:08 AM Christoph Hellwig <hch@xxxxxxxxxxxxx> wrote: > > On Fri, Oct 27, 2023 at 09:03:43AM +0300, Amir Goldstein wrote: > > With all due respect, your NACK is uncalled for. > > It abosolute is not. We must not spread the broken btrfs behavior. > > > The "details" that fanotify reports and has reported since circa v5.1 > > are the same details available to any unprivileged user via calls > > to statfs(2) and name_to_handle_at(2). > > Yes, and that's very broken. btrfs sneaked this in and we need to > fix it. > > Again: > > Extra-hard-Nacked-by: Christoph Hellwig <hch@xxxxxx> > > and I'm a little sad that you're even arguing for sneaking such broken > thing in. OK. You are blaming me for attempting to sneak in a broken feature and I have blamed you for trying take my patches hostage to promote your agenda. I hope that we are both wrong. Suppose that we follow up on your plan to fix btrfs, that it is all done and everyone is happy. You said it yourself, that we cannot risk regressions by changing the existing st_dev/f_fsid without an opt-in mount/mkfs option. Is that correct? If that is the case, fanotify will need to continue reporting the fsid's exactly as the user observes them on the legacy btrfs filesystems. The v2 patches I posted are required to make that possible. However, I think we can reach a compromise here. My main goal with the recent f_fsid series is to allow fanotify to be used as (unprivileged) inotify drop-in replacement. To do that, there is no need to allow btrfs sb/mount watch on sub-volumes, so I don't mind keeping the ban on sb/mount watches on btrfs sub-volume and relaxing only inode watch inside btrfs sub-volume. IOW, on legacy btrfs sub-volumes, users could use fanotify to watch changes/access to a file and get events with the same fsid that the user observed with statfs(2) on that file. Migrating to new btrfs sub-volumes (i.e. with unique sb/vfsmount) would enable the feature of fanotify sb/mount watch, so there is one more incentive for the community to fix btrfs fsid's and for users to actually make this migration. Will that address your concerns? Thanks, Amir.
