Hi Adrian, On Tue, 2025-04-22 at 14:35 +0200, John Paul Adrian Glaubitz wrote: > Hi Slava, > > On Mon, 2025-04-21 at 21:52 +0000, Viacheslav Dubeyko wrote: > > I am trying to elaborate the HFS/HFS+ maintainership action items: > > (1) We need to prepare a Linux kernel tree fork to collect patches. > > Yes. I suggest creating a tree on git.kernel.org. > Makes sense. Whom do we need to ask to make it happened? > > (2) I think it needs to prepare the list of current known issues (TODO list). > > Shall we use the kernel wiki for that? I suggest starting with the collection > of known CVEs as well as possible patches. I know of at least one CVE that > Ubuntu has fixed locally. > We can do this, but, as far as I can see, the kernel wiki's pages are marked as obsolete content. Also, Bugzilla could be more suitable for this. And, yes, we need to collect all known CVEs somewhere. > I can send an email to the author of that patch and ask them to send their > patch upstream. > Let's prepare the kernel tree and WiKi or/and Bugzilla at first. > From my memory, there are some occasional filesystem corruptions reported > on HFS partitions which might be a result of a bug in the kernel driver. > > They can be easily fixed with fsck_hfs from hfsprogs though. > Yeah, I need to check the email list for the issue reports. Even if some issues can be fixed by the fsck_hfs, potentially, some issues could be harmful enough. So, it makes sense to fix it. > > (3) Let me prepare environment and start to run xfstests for HFS/HFS+ (to check > > the current status). > > I suggest a Debian VM for that as it has hfsprogs which allows creating both > HFS and HFS+ filesystems. It's also easily possible to test on PowerPC inside > QEMU if necessary. > Sounds good! Do you mean a particular link with ready-made Debian VM images? > > (4) Which use-cases do we need to consider for regular testing? > > Definitely testing both legacy HFS and HFS+ with creating new filesystems, writing > and reading random files from it as well as running fsck on these. > > I'm not a Linux kernel filesystem expert, so I don't know what the recommend tests > for CI are, but I suggest everything that is commonly used, both with HFS and HFS+. > Makes sense. First of all, xfstests is a good tool for file system's functionality regression testing. Probably, I need to take a deeper look into fstests, finally. Maybe, we need to consider fio tool for the testing too. I think that it makes sense to check all supported logical block sizes and some set of volume sizes. I need to double check the HFS/HFS+ features, and we will need to test that supported features are not broken. And, of course, we will need to be sure that file system volume is consistent after been used by x86 and PowerPC platforms. So, let me spend some time for the testing strategy elaboration. > Thanks, Slava.
