Re: zfs and device name changes

[Takashi YAMAMOTO <yamt9999%gmail.com@localhost>]

hi,

On Tue, Mar 24, 2026 at 9:58 AM Simon Burge <simonb%netbsd.org@localhost> wrote:
>
> Hi Takashi,
>
> Takashi YAMAMOTO wrote:
>
> > the attached patch is my attempt to make zfs a bit more robust against
> > device name changes.
> > the identical patch is available at github too:
> > https://github.com/yamt/netbsd-src/commit/32283c2e362034301c3da218a05849c04ee20c2a
> >
> > while it seems working as far as i tested, i'd be happy if someone can review it
> > as my knowledge of zfs (well, and recent netbsd in general) is weak.
>
> There is some reasonable overlap with your patch and my work getting
> NetBSD booting from a ZFS root file system:
>
> https://github.com/snarkophilus/src/blob/zfsboot/external/cddl/osnet/dist/uts/common/fs/zfs/vdev_disk.c
>
> Your device_is_eligible_for_vdev() has overlap with what RAIDframe does.
> I introduced a "storagepool" concept which Chuck didn't really like:
>
> https://mail-index.netbsd.org/tech-kern/2022/08/12/msg028311.html

i wasn't aware of the effort. thank you.
do you want me to suspend my version?

i'm not sure if D_STORAGEPOOL concept is generic enough to teach every
drivers either.
for zfs, the actual requirement is something like just "large enough".
(cf. SPA_MINDEVSIZE)

if we take that approach, i guess it should be visible to userland. (eg. libzfs)
and probably for the corresponding cdev too.

>
>
> https://github.com/snarkophilus/src/tree/zfsboot is my working tree.
> I'm using this (with a few uncommitted mods) on my main server now.
>
> I need to find time to finish this work up, but alas I've been saying
> that for a quite while now.  The kernel and userland bits are mostly ok,
> but the bootloader is a complete mess as far as both code cleanliness
> and source file organisation.  I don't have a good feel for how long it
> would take to get that clean enough to commit.
>
> Now that you're looking at ZFS in more detail, I will try to make some
> movement on my ZFS changes.

(asking without looking at your code yet)
is it based on freebsd's one?
i hope it doesn't rely on spa_config_load, which has been retired in openzfs.

>
> Cheers,
> Simon.