NetBSD Problem Report #60938

From zakinko@snowrabbit.org  Thu Oct 29 17:17:03 2026
Return-Path: <zakinko@snowrabbit.org>
Received: from mail.netbsd.org (mail.netbsd.org [199.233.217.200])
	(using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits)
	 key-exchange X25519 server-signature RSA-PSS (2048 bits) server-digest SHA256
	 client-signature RSA-PSS (2048 bits) client-digest SHA256)
	(Client CN "mail.netbsd.org", Issuer "YR1" (not verified))
	by mollari.NetBSD.org (Postfix) with ESMTPS id A4A721A923B
	for <gnats-bugs@gnats.NetBSD.org>; Thu, 29 Oct 2026 17:17:03 +0000 (UTC)
Message-Id: <20261029171700.D5E95C9296@bambi.snowrabbit.org>
Date: Fri, 30 Oct 2026 02:17:00 +0900 (JST)
From: Showta Ishizaki <zakinko@snowrabbit.org>
To: gnats-bugs@NetBSD.org
Cc: zakinko@snowrabbit.org
Subject: makefs -o version=2 makes UFS2ea, unlike newfs -O 2
X-Send-Pr-Version: 3.95
X-From4GNATS: "Showta Ishizaki via gnats" <gnats-admin@NetBSD.org>

>Number:         60938
>Category:       bin
>Synopsis:       makefs -o version=2 makes UFS2ea, unlike newfs -O 2
>Confidential:   no
>Severity:       non-critical
>Priority:       low
>Responsible:    bin-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Thu Oct 29 17:20:00 +0000 2026
>Originator:     Showta Ishizaki
>Release:        NetBSD 11.0 and -current (11.99.9)
>Organization:
>Environment:
System: NetBSD 11.0 amd64
Architecture: x86_64
Machine: amd64

>Description:

I noticed this while booting a -current kernel from 2021 on a disk
image made with makefs -t ffs -o version=2: the kernel refused the root
file system with EFTYPE.  The image had magic 0x19012038, UFS2ea, and
not plain UFS2.

Since the UFS2ea split of 2022-11-17, newfs -O 2 makes plain UFS2 and
UFS2ea needs -O 2ea.  makefs got an extattr option in the same commit,
but its default is 1, so -o version=2 alone gives UFS2ea.  makefs(8)
lists the option without saying what the default is.

The 11.0 release images are made this way too.
distrib/amd64/installimage and liveimage set IMGFFSVERSION=2 and pass no
extattr, and the root partitions of NetBSD-11.0-amd64-install.img and
NetBSD-11.0-amd64-live.img both carry magic 0x19012038.  They boot fine,
since the 11.0 boot loader knows UFS2ea, but a kernel from before the
split cannot mount them (the 2021 one above gave EFTYPE).  alpha's
emuimage and the wii image use version 2 as well.

>How-To-Repeat:

	$ mkdir root && echo hello > root/f
	$ makefs -t ffs -s 8m -o version=2 a.img root
	$ od -A n -t x4 -j 9564 -N 4 a.img
	         19012038

With -o version=2,extattr=0 the same command gives 19540119.

>Fix:

Make the default 0, so that makefs and newfs agree, and say so in the
manual.  -o extattr=1 still gives UFS2ea for anyone who wants it.

I built makefs from -current (ac67e1168) on NetBSD 11.0/amd64 with and
without the diff.  With it, -o version=2 gives 19540119; extattr=0,
extattr=1 and version=1 give the same magic as before.  fsck_ffs -fn
from 11.0 is clean on every image.  The makefs steps of
tests/sbin/fsck_ffs/t_check_quotas.sh, the one test that runs makefs -o
version=2, give the expected quota messages with the new makefs.  I did
not mount the images (no root on that machine).

Diff against -current of 2026-10-30:

	--- usr.sbin/makefs/ffs.c
	+++ usr.sbin/makefs/ffs.c
	@@ -205,7 +205,7 @@ ffs_prep_opts(fsinfo_t *fsopts)
	 	ffs_opts->avgfilesize= -1;
	 	ffs_opts->avgfpdir= -1;
	 	ffs_opts->version = 1;
	-	ffs_opts->extattr = 1;
	+	ffs_opts->extattr = 0;	/* plain UFS2, as newfs -O 2 */

	 	fsopts->fs_specific = ffs_opts;
	 	fsopts->fs_options = copy_opts(ffs_options);
	--- usr.sbin/makefs/makefs.8
	+++ usr.sbin/makefs/makefs.8
	@@ -374,7 +374,7 @@ Range: 1..2147483647.
	 .It Cm extattr= Ns Ar num , Cm e= Ns Ar num
	 UFS2 with extended attributes.
	 One of:
	-0 to disable, or
	+0 to disable (default), or
	 1 to enable.
	 .It Cm extent= Ns Ar num , Cm x= Ns Ar num
	 Maximum extent size.

Home	PR Database Search