Announcement

Collapse
No announcement yet.

Linux 4.19.8 Released With BLK-MQ Fix To The Recent Data Corruption Bug

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Linux 4.19.8 Released With BLK-MQ Fix To The Recent Data Corruption Bug

    Phoronix: Linux 4.19.8 Released With BLK-MQ Fix To The Recent Data Corruption Bug

    Hopefully you can set aside some time this weekend to upgrade to Linux 4.19.8 as there's the BLK-MQ fix in place for the recent "EXT4 corruption issue" that was plaguing many users of Linux 4.19...

    http://www.phoronix.com/scan.php?pag....19.8-Released

  • #2
    So this is what 4.19.0 should have been? Sorry, not trying to hurt anyone's feelings, just pointing out the facts.

    Comment


    • #3
      The bug was not a concern for most people since you had to go out of the way to disable the I/O Scheduler, enable this blk-mq thing and so on, most distros ship with CFQ has the default scheduler, where this bug will not manifest. In most cases blk-mq is disabled by default, to tell if it is enabled, cat /sys/block/dm-X/dm/use_blk_mq, where X is 0 to 3.

      Also, cat /sys/block/{DEVICE-NAME}/queue/scheduler for each device shows what I/O scheduler you are using.

      Comment


      • #4
        Originally posted by Brisse View Post
        So this is what 4.19.0 should have been? Sorry, not trying to hurt anyone's feelings, just pointing out the facts.
        Yeah... because everyone and his mother knows that x.x.0 releases are completely free of bugs.

        Comment


        • #5
          Originally posted by jpg44 View Post
          ...In most cases blk-mq is disabled by default, to tell if it is enabled, cat /sys/block/dm-X/dm/use_blk_mq, where X is 0 to 3.
          https://bugzilla.kernel.org/show_bug.cgi?id=201685#c284

          Comment


          • #6
            Originally posted by jpg44 View Post
            The bug was not a concern for most people since you had to go out of the way to disable the I/O Scheduler, enable this blk-mq thing and so on, most distros ship with CFQ has the default scheduler, where this bug will not manifest. In most cases blk-mq is disabled by default, to tell if it is enabled, cat /sys/block/dm-X/dm/use_blk_mq, where X is 0 to 3.

            Also, cat /sys/block/{DEVICE-NAME}/queue/scheduler for each device shows what I/O scheduler you are using.
            Hm. I'm pretty sure I experienced this and all I had done was install 4.19.x using the UKUU tool on Linux Mint 19 Cinnamon edition (which is an Ubuntu 18.04 derivative). At first I just assumed it was SSD failure even though it's a lightly used Samsung 840 EVO, but it shows no other signs of failure so I was getting frustrated. Then I saw something about ext4 corruption in 4.19 and it was making a lot more sense. Judging by the posts of clueless users like me it looks like others had similar experience.

            I'd upgrade my kernel again to see if that scheduler was being used but I don't feel like borking my system again if that is the case. I did archive the filesystem before wiping the disk, would I be able to tell which scheduler was being used?

            Comment


            • #7
              Well, sounds like it is time to upload 4.19 into Debian

              Comment


              • #8
                Originally posted by dungeon View Post
                Well, sounds like it is time to upload 4.19 into Debian
                My wallet is ready.

                Oh, wait... That idiom doesn't quite work on free software does it? Bummer...

                Comment


                • #9
                  Originally posted by Overlordz View Post

                  Hm. I'm pretty sure I experienced this and all I had done was install 4.19.x using the UKUU tool on Linux Mint 19 Cinnamon edition (which is an Ubuntu 18.04 derivative). At first I just assumed it was SSD failure even though it's a lightly used Samsung 840 EVO, but it shows no other signs of failure so I was getting frustrated. Then I saw something about ext4 corruption in 4.19 and it was making a lot more sense. Judging by the posts of clueless users like me it looks like others had similar experience.

                  I'd upgrade my kernel again to see if that scheduler was being used but I don't feel like borking my system again if that is the case. I did archive the filesystem before wiping the disk, would I be able to tell which scheduler was being used?
                  Thats interesting. Unless you changed the settings, it would probably be using CFQ. Its kind of a hard setting to change so unlikely you would do it by accident. You are using an SSD, so maybe the distro might have turned blk-mq on but in the spinning disks I have seen the blk-mq thing is off, but just guessing here. It could be the bug is more widespread and affects more people who have not manipulated their scheduler settings. Regardless of what your configuration is, its a good idea to avoid the 4.19.0 kernel.

                  Comment


                  • #10
                    Originally posted by jpg44 View Post
                    The bug was not a concern for most people since you had to go out of the way to disable the I/O Scheduler, enable this blk-mq thing and so on, most distros ship with CFQ has the default scheduler, where this bug will not manifest.
                    Again proves that distros in general don't give a sh*t about performance.

                    Comment

                    Working...
                    X