FedoraForum.org - Fedora Support Forums and Community
Results 1 to 8 of 8
  1. #1
    Join Date
    Oct 2015
    Location
    Sheffield UK
    Posts
    85

    Help NVME Drive Resetting

    I've got a real problem at the moment, ever since upgrading to kernel 4.12 - my nvme drive has been resetting itself. I use this drive as my boot drive and because of this the root file system becomes read-only and eventually the system locks up.

    Not sure how to debug this one or gather enough information to file a bug report, as the logs aren't getting written to disk for obvious reason. Does anyone have any suggestions as to how I might capture what's going on?

    I don't believe this to be an issue with the drive (Samsung pm961) as it's been working fine for around 9 months and the smart data doesn't show an increased error count and/or any failed checks.

  2. #2
    Join Date
    Oct 2015
    Location
    Sheffield UK
    Posts
    85

    Re: Help NVME Drive Resetting

    My system is now completely unstable - this happens once a day, kernel panics are now occurring:


  3. #3
    Join Date
    Dec 2013
    Location
    United Kingdom
    Posts
    5,220

    Re: Help NVME Drive Resetting

    can you still boot into 4.11.11? If so, please report the 4.12 bug if you haven't already done so and use 4.11.11 for now

  4. #4
    Join Date
    Oct 2015
    Location
    Sheffield UK
    Posts
    85

    Re: Help NVME Drive Resetting

    4.11 exhibits the same problem - looks like it may be related to apst which was introduced for the nvme driver in 4.11.

    I've filed a bug and also written a patch to disable aspt deep sleep for this drive, which is in the report. I'm in the process of rolling my own kernel now - fingers crossed.

    https://bugzilla.redhat.com/show_bug.cgi?id=1487421

  5. #5
    Join Date
    Oct 2015
    Location
    Sheffield UK
    Posts
    85

    Re: Help NVME Drive Resetting

    The plot thickens -

    Looks like there are specific quirks in the kernel for some similar drives, but only when accompanied by specific motherboards. I'm in a bit of a mess, because also built a production system around several of these drives with rhel - really hope this bug doesn't filter down.

    An another note I made a mistake with my original patch - made a typo (late night), I've now reposted the patch and have a running kernel. No resets so far - fingers crossed, I really wish I could get some acknowledgement from redhat on the bug report.

  6. #6
    Join Date
    Dec 2013
    Location
    United Kingdom
    Posts
    5,220

    Re: Help NVME Drive Resetting

    Hopefully it will be a good few years before this 4.12 kernel branch hits RHEL by which time such bugs will be ironed out.

    Even for Fedora I reckon the kernel was pushed stable far too early as the bugs that have occurred since updating to it are not something users can put up with 'for the greater good'. I commented (about the same length of this post) about this mess in one of my own filed bugs and got asked not to rant because it isn't helpful. If I was going to rant, it would have been much longer, and nowhere near as restrained language wise. What I posted was the most polite way of getting my frustration across instead of going all out Linus Torvalds "&*() you NVIDIA!" on them

  7. #7
    Join Date
    Oct 2015
    Location
    Sheffield UK
    Posts
    85

    Re: Help NVME Drive Resetting

    I can confirm that the kernel patch that I wrote has solved the issue for me.

    Unfortunately this affects any kernel >= 4.11 - essentially wherever apst is enabled on the nvme driver. Looks to be a known issue with the apst code that means it interacts poorly with some Samsung firmware versions.

    Earlier kernels aren't available on Fedora 26 - what do I do? I'm not getting any response from redhat - despite providing the solution. Am I stuck compiling my own kernel?

  8. #8
    Join Date
    Oct 2015
    Location
    Sheffield UK
    Posts
    85

    Re: Help NVME Drive Resetting

    Just pinged this over to the mailing list...

    They suggested that I need to submit this patch upstream - seriously? I'm not a kernel developer - I'll most likely get eaten alive.

    Arghhh.

Similar Threads

  1. [SOLVED] Samsung NVMe
    By theking2 in forum Hardware & Laptops
    Replies: 7
    Last Post: 16th October 2017, 10:50 PM
  2. Replies: 3
    Last Post: 25th July 2017, 10:29 AM
  3. Any two port NVMe PCIe SSD adapters out there?
    By ToddAndMargo in forum Hardware & Laptops
    Replies: 0
    Last Post: 24th May 2017, 08:24 AM
  4. PCI and nvme errors
    By RobertAdesam in forum Hardware & Laptops
    Replies: 2
    Last Post: 25th January 2017, 10:39 AM
  5. gparted doesn't see Intel 750 NVME SSD (Fedora 22)
    By bjrosen in forum Hardware & Laptops
    Replies: 1
    Last Post: 16th September 2015, 01:38 AM

Tags for this Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •