Date: Wed, 23 Mar 2022 16:26:16 -0600 From: Alan Somers <asomers@freebsd.org> To: freebsd-fs <freebsd-fs@freebsd.org> Subject: What's the best FS-agnostic test suite for a file system's data path operations? Message-ID: <CAOtMX2hJ_hnWbgvoivTB4yR3AhC5RKEeE1xHiw=kbYq5yjE=kQ@mail.gmail.com>
next in thread | raw e-mail | index | archive | help
pjdfstest is great at testing control-path operations, like mkdir, chmod, etc. And it works on any POSIX-compliant file system. But it doesn't include any datapath operations like write, read, or copy_file_range. fsx is great for finding datapath bugs and it also works on any posix-compliant operating system. However, it's a random torture test, not deterministic. Does anybody have any good ideas for how to either turn fsx deterministic, or else a good argument that random torture tests are inherently superior to deterministic tests for datapath operations? Alternatively, does anybody know of a good alternative to fsx? I've heard of fsstress and xfstests. But the former looks pretty much like an fsx clone, and the latter isn't totally file-system agnostic. There's some urgency here: I've already had five inquiries about my pjdfstest rewrite proposal for GSoC. I hate turning people down, so I'm wondering if I can extend the proposal to accomodate two students, or maybe create a second proposal. Some possibilities: * Merge fsx into pjdfstest (my first idea) * Extend fsx to include newer operations like copy_file_range and fspacectl (probably too small for a whole summer) Another GSoC idea would be to implement virtio-fs. However, I haven't done my own homework on that yet. I don't know how hard it would be. If anybody has any other ideas, please add them to the wiki at https://wiki.freebsd.org/SummerOfCodeIdeas ASAP because there's a lot of interest this year. -Alan
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAOtMX2hJ_hnWbgvoivTB4yR3AhC5RKEeE1xHiw=kbYq5yjE=kQ>