Re: diagnosing post-install boot failure

[Ben Collver <bencollver%riseup.net@localhost>]

On Wed, Feb 25, 2026 at 01:07:29PM +0100, Martin Neitzel wrote:
> BC> dmesg shows intelfb0 at i916drmkms0
> 
> Do you also happen to have dmesgs refering to the state of the
> EDID recognition, i.e.  the resolution and timings of your
> monitor/display?

intelfb0: framebuffer at 0x90040000, size 1360x768, depth 32,
stride 5440

> I'm asking because NetBSD will behave the same on an old
> ASUS EeePC 1003H (Intel N270 CPU, 945GME graphics):
> 
> It will successfully probe for the i915drmkms, and then,
> when switching from VGA/BIOS to the framebuffer, just
> turn black.

This sounds like what i was seeing.

A friend gave me an Intel NUC7i3BNH manufactured in 2018,
and i am trying to get NetBSD running well on it.

For many months it reliably ran Alpine Linux (no X11), with one
caveat. Sometimes the boot process took an amazingly long time.
Recently i fixed that by replacing the coin cell battery,
which was not simple.  This model has a connector with leads
"soldered" onto the battery and covered with shrink-wrap.  I removed
the shrink-wrap with a utility knife, pulled the leads off the dead
battery, and reconnected them to a new battery using tape.  It booted
more quickly after that.

> Disabling the i915drmkms driver (via "boot -c" / /boot.cfg)
> helped me to get started (as others have already recommended).
> (In the older days, the pre-KMS "i1915drm" would take over
> and do a good job, too;  but I believe we have dropped that
> driver by now, at least in the default GENERIC kernel.)

Maybe i'll give that driver a try...

> With an installed system, you can give i915drmkms another
> try and then, with a dark/unusable console ssh into the box
> and check for the dmesg output.

Unfortunately, this morning i can no longer reproduce this problem.
I guess it was intermittent, and it may have been a coincidence
that it seemed to clear up when i de-selected the gpufw set during
install.

I also experienced another intermittent problem.  At random i got the
following kernel error:

    iwm0: fatal firmware error

After this error, WIFI would stop working until i power cycled the
system. I plugged in a Realtek USB WIFI adapter and that seems to be
more reliable.

> In my EeePc case, I found a complaint the EDID checksum failing
> (and hence the EDID data being ignored).  None of the other
> OSes (*BSD, Linuxens) had a problem with that.  I simply
> commented out that checksum check in my sources and it happens
> to work great:  I can boot using the i915drmkms driver, and
> it switches to a high resolution, usable framebuffer console,
> and X11 will work, too.

Interesting detail.  I don't see the word "checksum" in the dmesg
output.  Thanks for the wscons tips.

Best regards,
-Ben