Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror
×
Linux

ReiserFS Proposed To Be Removed From Linux In 2022 (phoronix.com) 217

UnknowingFool writes: Linux kernel developers have discussed on the kernel developers forum to remove ReiserFS from the kernel starting in 2022. ReiserFS was added as Linux's first journaling file system 21 years ago with SUSE using it as the default filesystem until 2006. However, since Hans Reiser was sent to jail 15 years ago for murder, there has not been much development or interest in it. Noting that there have been no user-spotted fixes since 2019, longtime kernel developer Matthew Wilcox also cited that ReiserFS was only block for some kernel changes he wished to implement. These days there are better alternatives like EXT4, Btrfs, XFS, and OpenZFS.
This discussion has been archived. No new comments can be posted.

ReiserFS Proposed To Be Removed From Linux In 2022

Comments Filter:
  • by sien ( 35268 ) on Wednesday February 23, 2022 @05:14AM (#62294747) Homepage
    Alas this file system comparison page from wikipedia [wikipedia.org] will no longer be relevant.
    • by Ecuador ( 740021 )

      Why not? It will not lose its status as a "file system".

      I remember ReiserFS speeding up a system that had millions of tiny files considerably back then, it was very good at the time. I guess if there was no murdering or no catching the murderer, it might still be relevant in some current version.
      I am not sure how to feel about it, I mean, on the one hand, if I only used software made by ethical people, would I have any software to use? On the other hand, there are "tiers" of unethical/criminal activity, so

      • The fact the file system is named after him is sort of a red flag. Makes it hard to put on your resume.

        • Nothing stopping anyone from forking the code and rebranding it.

          But what compelling reason would there be to choose NinaFS in 2022 over filesystems that have been battled-tested through support and development over the past decade and a half within Linus' official tree?

          • by jythie ( 914043 )
            Even at the time of the trial, if the FS had enough utility it probably would have been rebranded or at least have people step up and develop it further. Often in science and technology, there is a balancing act when it comes to taint vs utility, where if something has a bad reputation because of the people involved, humans will find a way to work around that if something is useful enough. ReiserFS had some utility, but it was easier to learn from it and build other better systems than to maintain the cod
      • by shabble ( 90296 ) <metnysr_slashdot@shabble.co.uk> on Wednesday February 23, 2022 @06:35AM (#62294847)

        Why not?

        Have a closer look at the last column on that table.

        • by Ecuador ( 740021 )

          Well, yes, that's what we are talking about. GP said the file system comparison will no longer be relevant. I am asking why not, it is still going to be in a list of "file system" after it's removed from the kernel, and I haven't heard anything about his wife being unmurdered, so the table will still be relevant...

      • by loonycyborg ( 1262242 ) on Wednesday February 23, 2022 @06:35AM (#62294849)
        It's still developed as reiser4-5, but those iterations weren't merged into kernel for whatever reason. What is currently in kernel is reiser3 which is obsolete even from point of view of original developers.
        • Reiser4-5 is considered a separate branch, and it has not been added to the kernel due to lack of support. Reiser4 or 5 could be added in the future however the current kernel branch is 3.6 and as you mentioned obsolete.
      • by jd ( 1658 )

        I used ReiserFS for the /tmp filesystem for precisely that reason. Most files there are lock files or small workspace files, so having something that accelerates that case is actually very useful. However, as /tmp is wiped on reboot, it would be better to have a filesystem that handles that case which doesn't journal, as that's data you will never be replaying. I can therefore see a more stable, trimmed-down ReiserFS as having value. But that would be a new FS, so would have a different name.

        There may be be

  • Surely the question is whether there are any known bugs that have been left unpatched. What is a "user-spotted fix" anyway?

    • by Entrope ( 68843 )

      Fixes come from three sources: users, people who care about that particular code (developers and maintainers), and people who are mostly interested in other things but encounter that code in the course of what they care more about.

      For reiserfs, the second group is currently empty. So if users are not finding bugs, either the code is super mature, or the maintenance burden falls on the third group of people (who could be spending the time instead on other things). Linux does not try to maintain internal AP

      • by nagora ( 177841 )

        So, it's really "user-supplied fixes", then?

        • by Entrope ( 68843 )

          If you take "user-spotted bugs" literally, that is even worse: There is not enough "organic" use of the filesystem to find bugs, so it takes time from syzkaller and similar teams to even exercise the code, and then somebody else has to take the time to analyze and fix the failure.

      • by trparky ( 846769 )

        Linux does not try to maintain internal API stability,

        I've often wondered why that is. Can someone tell me why?

        • by Jeremi ( 14640 )

          I've often wondered why that is. Can someone tell me why?

          Presumably it's the usual reason -- they decided that the overhead of maintaining backwards compatibility (in terms of additional testing that would need to be done, additional size and complexity added to the codebase, and the limitations it would place on design improvements that could be made) was too large to be worthwhile, and they'd rather suffer the pain of having to update the calling code than have to support "the old ways" indefinitely.

    • by jd ( 1658 )

      There are a lot of known bugs, although quite a lot of them are untriaged.

  • by DrMrLordX ( 559371 ) on Wednesday February 23, 2022 @05:32AM (#62294767)

    Paul is appalled.

  • by The Evil Atheist ( 2484676 ) on Wednesday February 23, 2022 @05:58AM (#62294793)
    They're washing their Hans of it.
  • by Gabest ( 852807 ) on Wednesday February 23, 2022 @07:08AM (#62294881)

    He had all that time to develope it to perfection as his rehabilitation. He could have been a useful to society. What a waste!

  • There was some merit to using this file system back in the day. What was better than the file system was the daily lkml chatter about the system and the inventor. It could have been a miniseries.
  • by Anonymous Coward on Wednesday February 23, 2022 @07:19AM (#62294915)
  • by ffkom ( 3519199 ) on Wednesday February 23, 2022 @08:47AM (#62295115)
    One of reiserfs' unique selling points is its ability to store small files efficiently, by storing their (shorter-than-one-full-block) endings together, instead of just wasting the rest of a full block.

    Which of the proposed alternatives is as efficient as reiserfs when storing millions of little files?
    • by _merlin ( 160982 )

      NTFS can store files smaller than a block in the directory entry itself. Efficient storage for small files isn't exactly a high-tech feature these days.

    • While this feature could be useful for older machines with less storage and lots of small files, is it useful for modern systems with GB sized files and TB space? From what can tell this feature would be good for KB sized files on MB HDDs.
      • by jd ( 1658 )

        /tmp is still mostly lock files, so basically files of length 0-8 bytes (depending on whether a PID was with them).

  • So.. I'm one of those people who does use reiserfs. If you send me an email, it's stored in that. That's all I use it for, but that's something.

    What other filesystems perform as well (or at least within an order of magnitude, no more than ten times slower) for a situation where you have tens of thousands of little files all in the same directory?

    Or do people simply not do Maildir anymore (the idea being that if you have tens of thousands of files in a directory, then "you're holding your phone wrong")?

    • Food for thought, but what about NTFS? The performance gains in 5.15 are tremendous, but I don't know if a proper evaluation has been done, particularly in your use case. I know that culturally, NTFS dates back to the mid-90's and was therefore built around efficiently packing and accessing small files and data, and it's been extended to support attributes that didn't exist, including support for Linux-style file permissions in order to facilitate better Linux compatibility under the WSL environment.

      5.15
    • For just one mailbox? It might feel good to know that you're getting the most performance possible, but will you really be able to tell the difference just moving it to Btrfs or EXT4? Unless it's an 8086 with a dying hard drive, it's probably going to be nearly imperceptible.

  • Don't really know enough about the relative merits of ReiserFS over others, but just because the guy murdered his wife shouldn't impact the quality, or lack thereof, of his code. They could just fork and rename it to something else if they're worried about the name association.

    • by ledow ( 319597 )

      The prime (almost sole) developer and architect of the filesystem is not in a position to maintain it, and that's unlikely to change in the foreseeable future.

    • I think you missed the "lack of interest" and "there are better alternatives" bits.

    • by jythie ( 914043 )
      At the end of the day, the merits were not enough. If it had enough utility, as you say, someone would have forked it, or at least kept working on it. While they were paid to do so, other developers had been working on it for quite some time, and any of them could have taken over either for themselves or it could have been picked up by any number of companies that saw value in it. The file system space is already a bit cluttered, and other really good systems existed and continue to exist. ReiserFS had
  • ReiserFS is a killer file system. :)

The question of whether computers can think is just like the question of whether submarines can swim. -- Edsger W. Dijkstra

Working...