Running FreeBSD on M.2 SSD

Mario Olofo mario.olofo at gmail.com
Mon Feb 24 17:58:26 UTC 2020


Hello John, thank you for your reply.

Yesterday I reinstalled the 12.1 on a VirtualBox virtual machine, did the
same steps and it didn't corrupted the ZFS, so I think that the problem is
in the FreeBSD's driver for m.2 SSD.
Besides the corruption of the filesystem, I forgot to mention that I
noticed a little noise on disk writes on FreeBSD, but not on Linux or
Windows.
I found some old threads about incorrectly params for sector size for
Samsung's SSD, but nothing about WD.
If someone responsible for the driver need help to solve this problem, I
can reinstall the FreeBSD on my machine and compile a custom kernel to
gather debug information.

Thank you,

Mario

Em seg., 24 de fev. de 2020 às 11:47, John Kennedy <warlock at phouka.net>
escreveu:

> On Sun, Feb 23, 2020 at 11:18:08PM -0300, Mario Olofo wrote:
> > Some time ago I tried to switch from Linux to FreeBSD 12.1, used a WiFi
> > dongle and all good, until I found that both ZFS and UFS corrupted the
> > filesystem very fast.
> > I work with a lot of small files because of web programming
> (node_modules),
> > so after a clean install, after installing the dependencies for my
> project,
> > if I scrub the zpool, it always found that the system is corrupted and
> > never recover.
> >
> > I have a WD Green M.2 SSD 480GB WDS480G2G0B.
> > Both Linux and Windows work correctly and don't detect any problems with
> > the disk.
> >
> > Did someone knows if it isn't supported by FreeBSD or there's some
> specific
> > configuration params that I need to set to it work correctly?
> >
> > I made a post on the forums back in the day I had the problem, the logs I
> > had are all there:
> >
> https://forums.freebsd.org/threads/fixing-metadata-errors-after-zfs-clear-zfs-scrub.72139/
>
>
>   Can't answer your WD Green question specifically, but I'm happy with my
> setup, below.  Good to look for quirks, but you probably also want to list
> other hardware involved as well (which might have it's own quirks).  If
> you've
> had good success (and no corruption) with two other operating systems on
> the
> same hardware, I'd probably be looking at software and/or drivers, and that
> requires knowledge of the hardware.
>
>
>   I've got dual EVOs (below is just from one I'm typing on) on two
> different
> FreeBSD boxes.  Nothing specific I had to do in FreeBSD, although on the
> other
> motherboard I had to tweak the motherboard settings to give it the
> channels it
> needed to shine.
>
>         kernel: nvd0: <Samsung SSD 970 EVO Plus 500GB> NVMe namespace
>         kernel: nvd0: 476940MB (976773168 512 byte sectors)
>         kernel: nvd1: <Samsung SSD 970 EVO Plus 500GB> NVMe namespace
>         kernel: nvd1: 476940MB (976773168 512 byte sectors)
>
>   If compiling kernel and packages from source count as having lots of
> little
> files, then I do as well.  I think I'm ZFS everywhere (boot partition being
> the question over time).
>
>   Personally, the only ZFS corruption I've had over time has been caused by
> bad hardware.  When I moved the disks to another box, they were fine with
> the same version of FreeBSD.  I scrub my zpool about once a month just
> because,
> plus after I get the kernel to crash.
>
>   The original box went all the way back to root-on-ZFS + FreeBSD 11.  The
> newer box just started around 12.0 (2019-05-31).
> _______________________________________________
> freebsd-stable at freebsd.org mailing list
> https://lists.freebsd.org/mailman/listinfo/freebsd-stable
> To unsubscribe, send any mail to "freebsd-stable-unsubscribe at freebsd.org"
>


More information about the freebsd-stable mailing list