The HDFS protocol supports POSIX.1e style ACLs. Supporting such ACLs in a multiprotocol environment means a translation method should be defined to translate between the NFSv3 mode bits, NFSv4 ACLs, Windows Style ACLs and the POSIX.1e ACLs. POSIX.1e ACLs differ in their structure and evaluation algorithm as compared to other ACLs that are currently supported in OneFS. The talk will detail the approach we took and also some of the surprising challenges related to multithreading.

Fast, scalable parallel file system performance is a key enabler of massively parallel computing as well as of emerging big data and machine learning applications. Released almost two decades ago, Lustre has long been the storage solution of choice for many supercomputing data centers. But as the world slowly retire rotational disks in favor of fast SSDs and persistent memory for their performance tiers, Lustre is becoming increasingly unable to fully utilize available storage bandwidth due to its old, disk-oriented object storage server designs based on Ext4 derived ldiskfs.

The UNIX Filesystem API is profoundly broken, and user-settable symbolic links are to blame. In this talk I will explain how CVE-2021-20316 made me realize how symbolic links, introduced in 4.2BSD Unix from U.C. Berkeley, broke the previously elegant UNIX file system API and file system design. The design and implementation of symbolic links has caused years worth of security flaws and API patches to fix a conceptually broken idea. I also propose a modest suggestion in order to help Linux step away from this mess to a more secure by-design future.