This is good, but it's a little confusing when reading the code because it isn't clear how we could get into this scenario. We could add a comment (similar to what you described in the message above), but it might be more clear to solve this with a change to a more relevant code path:
In zfs_rezget(), we could treat `z_links == 0` the same as if the file's deletion had completed on the source, and the object is no longer allocate (which is also the same as `gen != z_gen`), i.e. not reassociate this znode with the underlying object. Then if we try to access this file, we'll get EIO (see ZFS_VERIFY_ZP()). In addition to making it a little easier to understand the code, this also means that you couldn't see a partially-deleted file on the target (e.g. if we snapshotted this fs while the object is in the middle of being incrementally destroyed by dmu_free_long_range). We could change the code in zfs_zinactive() to an assertion, with a comment explaining why we can't see unlinked objects on a readonly mount. -- You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub: https://github.com/openzfs/openzfs/pull/464#issuecomment-328889887 ------------------------------------------ openzfs-developer Archives: https://openzfs.topicbox.com/groups/developer/discussions/Tf5d8079d50248fc8-M6560954ec3ae5ac8cadde41b Powered by Topicbox: https://topicbox.com
