[TUHS] the device tree, hardware, and kernels.

[Warner Losh via TUHS]

On Mon, Apr 13, 2026 at 10:26 PM Chris Hanson via TUHS <tuhs at tuhs.org>
wrote:

> On Apr 6, 2026, at 4:12 PM, segaloco via TUHS <tuhs at tuhs.org> wrote:
> >
> > DTB to me is better than having to go twiddling baked in kernel
> addresses at build time, then you really do need a kernel-per-machine
> rather than being able to spin a "defconfig" kernel that'll hopefully at
> least give you serial TTY on whatever you're messing with.
>
> This is the main reason I think someone is working on improving device
> tree adoption in NetBSD; availability of community and vendor device trees
> could reduce the number of more specialized kernel builds needed, and
> reduce the requirement to do a full port to support new hardware.
>

Yes. FreeBSD has been doing the same thing for a long time. Years ago, we
switched to primarily using DTS for ARM systems, while also including ACPI
support so we could get system information from either or both.

DTS are, in theory, a pure description of the hardware that any OS can
implement. Sadly, the reality has fallen short of this ideal in many areas
of the DTS world. Often times, the DTS and the Linux driver co-evolve (and
change year by year), which makes it hard to keep up. It also tends to
enforce a policy of updating everything (at least within an arch) at once
(it's just easier that way). However, this approach causes churn and
reduces longevity. For example, parts of our TI ARMv7 support were broken
for years before we removed them, and bringing the other parts back up to
standard required many hours. Overall, it's been great, but it hasn't been
without cost. Though armv7 supporters in FreeBSD have also been thinning
out too.

Warner