A lot of context in openshift/os#1744.
Summarizing AIUI: currently NFS exports in an image mode system where composefs is the root of the filesystem doesn't work unless either (1) the overlayfs has the nfs_export flag on, or (2) the exports file uses fsid=0 to set the "root share" somewhere else. The base issue is that the NFS kernel bits still need to be able to traverse from the overlayfs root to get to the actual exports, even if you're not interested in exporting things directly on the composefs overlay.
Should we just turn on nfs_export by default? As Colin mentioned in openshift/os#1744 (comment), it conflicts with metacopy=on which we do use but maybe don't actually need.
If not by default for whatever reason, it seems reasonable to at least have a mount options flag for it and libostree could either use that flag by default, or make that configurable in e.g. prepare-root.conf.
A lot of context in openshift/os#1744.
Summarizing AIUI: currently NFS exports in an image mode system where composefs is the root of the filesystem doesn't work unless either (1) the overlayfs has the
nfs_exportflag on, or (2) the exports file usesfsid=0to set the "root share" somewhere else. The base issue is that the NFS kernel bits still need to be able to traverse from the overlayfs root to get to the actual exports, even if you're not interested in exporting things directly on the composefs overlay.Should we just turn on
nfs_exportby default? As Colin mentioned in openshift/os#1744 (comment), it conflicts withmetacopy=onwhich we do use but maybe don't actually need.If not by default for whatever reason, it seems reasonable to at least have a mount options flag for it and libostree could either use that flag by default, or make that configurable in e.g.
prepare-root.conf.