mkimage: fix grammar in comments

Also generate hashes for files in gzimg directories.

At one point this should probably be unified at least a bit with the hash
generation of regular sets, so that there is a single place that defines
the types of hashes generated for sets and gzimg.

Fix resize_root.

Revision 1.85 added the 'log' option to "normal" fstabs, which is
incompatible with resize_ffs. Restore the previous behaviour.

mkimage: Allow configs to override fsize/bsize/ffs version

Inject TOOL_SED and TOOL_PWD_MKDB into mkimage

The configuration files used for mkimage might want to customize the passwd
database of the generated image (as is the case for an out-of-tree config
file I have) and passing in these tools helps to execute those modifications.

Recognize MKPOSTFIX=no in mkimage

avoid extra comma, add braces

Remove duplicated code, handle minwrites in _normal the same way we handle
it in _gpt. Initial patch from jmmv@.

Undo accidental commit.

Import Linux 6.10.3 DTS for RISC-V

This has the side effect of pulling in 5 arm64 renesas and a few other
dt-bindings files which didn't exist before so there are no conflicts.

Hopefully arm/arm64/mips can catch up soon.

branches: 1.82.2;
Add new sets: base32, debug32, manhtml

- base32 contains (when MKCOMPAT=yes) shared libraries for 32-bit
compatibility, previously included in base

- debug32 contains (when MKCOMPAT=yes) debug symbols and static libraries
containing debug symbols for 32-bit compatiblity, previously included
in debug

- manhtml contains (when MKHTML=yes) the HTML files previously included
in 'man', which are of limited utility without third-party software.

The motivation for this change is to be able to easily exclude sets
from CD-ROM images that go over the size limit without xz compression
(which many NetBSD platforms struggle to extract at acceptable speeds).

Quote argument to -n as required. [ -n ] is not a syntax error but
evaluates to true.

PR/57925: Jan-Benedict Glaw: Honor $MKREPRO_TIMESTAMP, don't embed name and
stamp in gzip.

evbppc: Add initial support for the Nintendo Wii

branches: 1.78.2;
Move DRM-driver firmware from base to its own set, gpufw

This set is only installed on amd64,i386,evbarm.
This set is installed on minimal installs and on install media, in
case someone needs it for basic driver functionality.

Comments:
Switched to a single MK tunable for it - that is probably unneeded.

An upcoming DRM update will include even fatter firmware, and we'd
like to minimize the impact of it.

Fix fallout from mkimage rev 1.76.

For mkimage:

- Update "size" if auto-calculated.
- Use "dd bs=1" instead of non-portable "head -c".
- Some style nits.

For MD make_label() functions:

- Stop using "newsize" as image size in MB, use "size" instead.

embedded/mkimage: Pad image with zeros to multiple of 4 MB plus 1 MB.

Otherwise, there may not be enough space after the ffs partition for a
gpt, leading to very confusing results.

Create an "octeon.img.gz" bootable image for evbmips64-eb release builds.

Add support for hybrid MBR/GPT images.

Add GPT support to mkimage.

Separate devicetree .dts -> .dtb building from kernel builds. They are now
part of a separate set, "dtb.tgz", and only built when MKDTB=yes. This
defaults to yes for earmv[67]* and aarch64, and no everywhere else.

Derive the FAT type from the MBR partition type and pass it on to makefs
when crating a msdos file system.

Split out /rescue to its own set and adapt installers/images to add it.

This is meant to make updates safer: if something goes wrong with updating
base, we still have the old, standalone /rescue to recover from.

branches: 1.69.4; 1.69.6;

Be more precise about exactly what fails when something does.

Relying upon set -e to abort things is sort of OK (it is not
a recommended option to use in general - too many odd special cases),
but only if user can work out from the "build failed" what actually
went wrong.

Tested only on amd64 build (for this, i386 is the same) - if anyone
has problems on builds for other systems, please let me know. However
the changes affect only failure paths, the most likely problem would
be for a build to fail to halt on an error, and I hope I have avoided
that. There should be no difference at all to error-free builds.

Handle use of TMPDIR in cleanup as well.

Honour ${TMPDIR} if set for location of image, rather than always
simply using /tmp

Replace HUMMINGBIRD_A31 with SUNXI kernel on armv7.img and include .dtb
files for SUNXI and TEGRA kernels on the MSDOS partition.

add the tests set.

branches: 1.64.4;
Now that FAT+FFS images include the BSD partition in the MBR partition
table, install the disklabel after setting up MBR partitions. This moves
the disklabel to the BSD partition from the start of the disk, avoiding
a conflict with the ODROID-C1 bootloader.

Improvements and fixes for FAT+FFS images:
- Add a "NETBSD" volume label to the MSDOS file-system.
- Explicitly initialize boot sector.
- Correct an issue where the MSDOS file-system was larger than the
partition table entry.
- Add NetBSD partition to the MBR partition table.

Mark the msdos parition active for boards like the BeagleBone Black

branches: 1.61.2; 1.61.4;
use a larger (64KB) block size, this helps SD card performance

Add a new endianes flag to mkimage and pass it on big endian arm builds.
From Rin Okuyama.

fix mkubootimage

Add some documentation:
Tell what this does, and what it expects from conf files

Properly spell X' name. From X(7):

The X.Org Foundation requests that the following names be used when
referring to this software:

X
X Window System
X Version 11
X Window System, Version 11
X11

initialize resize before pulling in board config

Add support for auto-growing the root partition. Enable it for rpi.img.

resizing of ffsv2 still doesn't work well, so revert previous, and instead
explicitly set fsize=2048/bsize=16384, which gives a pretty decent boost
to resize_ffs performance on its own

Use FFSv2 instead of FFSv1 for the root file-system. This should fix the
"resize_ffs takes forever" issues with evbarm images -- growing a v2
file-system is much faster than v1.

Fix previous so that the new generated rc.conf is picked up and that
any new new mount points (e.g. /proc and /kern) are generated.

While here increate the fs size so that postfix can actually build
/etc/mail/aliases.db

PR/49596: mkimage: incorrect permissions on evbarm images

Use NetBSD.dist in our mtree specification to populate the top level
directories, etc.

For each generated file hand craft an mtree entry.

Remove specialdirs as NetBSD.dist creates /proc and /kern.

Use [dx]sets_p as flag for debug/X sets inclusion. xsets is the list of
sets.

fix image building:
- round up the total size of the disk image
- fix fdisk offset for netbsd ffs images

branches: 1.48.4;
now that fdisk has optional arguments, don't pass -1

provide a -r rootdev option to handle the wd/sd/ld lossage.

these changes produce a bootable system under qemu, but its root is trashed.

Add preliminary support for creating x86 usb sticks with mkimage.
This is untested, and just a checkpoint of the work. Image looks
ok, but I have not tried if it boots.

Fix build release of evbarm which had been failing in mkimage
beagleboard_smp with the elusive message:

"gzip: can't stat: gzip: No such file or directory"

This was due to GZIP=gzip being defined in mkimage; we now use:

GZIP_CMD=${TOOL_GZIP:-gzip} # ${GZIP} is special to gzip(1)

as distrib/sets/sets.subr already does in these files as well:

distrib/utils/embedded/mkimage
distrib/utils/embedded/conf/beagleboard.conf
distrib/utils/embedded/conf/rpi.conf
distrib/utils/embedded/conf/rpi_inst.conf

SSD cards is stupid.

explain -m better

tidy up, fix usage

Use TOOL_GZIP in mkimage, and pass the TOOL_GZIP into it.

- explanations for single letter flags
- flag to add debug sets
- flag to minimize writes to sd's

Double extra to fix image size to fix build.

For ffs, set bytes per inode to 4K. Otherwise we get a density based on
the contents of the image, which aren't very useful after the file-system
has been resized -- for rpi.img, this was 22K per inode!

Add a new installer image for Raspberry PI to evbarm

Includes the following major changes:

1) Add new configmenu item to sysinst to allow creation of users at
post-install time.
2) Add an sshramdisk type to evbarm, which is a ramdisk with sshd
enabled, allowing users to install on a headless PI by sshing to the
device with the account "sysinst" password "netbsd" to begin the
installation. (Note, neither the account, nor the ssh setup are copied to
the finalized installation image)
3) Change mkimage to build a boot-only image, without the root filesystems.

Much of the sshramdisk code could be reused on other ports easily.

don't use [ -f $i ] && cat $i, since we are using -e

add extra space to root partition

set -e

set src= before pulling in host config

no sh -x

fix msdos mbr partition size calculation

don't dup -F

add -F to fdisk

mass -N db to makefs too

specify a database for mtree.

dont call mtree directly

really calling disklabel with the arguments of fdisk does not work.

- remove dup trap
- compare size to 0 for autosizing
- remove .gz file before compressing.

add compression support.

add tool hooks.

fix usage again -- -D specifies destdir not releasedir
include conf/${h}.conf not conf/rpi.conf
fix ./mkimage: line 110: "7" - 1 : syntax error: operand expected (error token is ""7" - 1 ")

fix usage

fix options

switch to makefs

Add the tests sets by default. Tweak the usage output.

More usage updates

Update usage output a little

- Add support for xsets
- Factor out common evbarm code

tidy up some more by factoring out the sizing code.

Make autosizing really work:
- Use newfs to compute the actual filesystem required size.
- Fix the computation of set sizes to account for blocks and fragments.
This results into a 95% full filesystem for my test (Total 566MB, 24MB free)
from which 8MB is my requested overhead. Not perfect, but good enough.

first step at autosizing, still does not work.

expand set name to path when checking to see if the tgz exists

move 'trap' after command args are parsed

honor ${TMPDIR}

complain on unknown options

move configuration to new place.

Adjust mkimage to handle RPI. From skrll@

branches: 1.3.2; 1.3.4;
consolidate all of the volatile /var.* directories which are mounted as tmpfs
into a single one for /var - same as /root, /etc

While the resulting image from mkimage can always be mounted with vnd,
and then modified, it is much easier to let this late, user-oriented
customisation be done automatically. These mods let mkimage change
information based on files in a "custom" directory, which can be set
in mkimage using -c customdir.

After all other configuration, the files under custom/ will be copied
into the main image. The files under custom/ are taken into account
when performing autosizing.

Also, generalise the sets directory to point to the current machine
architecture, although this can be overwritten from the command line
by anyone making cross-images.

An example to make a user image with an embedded read-only 100 MB
iSCSI target:

% ./mkimage -c custom
=== making a new 272 MB image in usermode.img ===
272+0 records in
272+0 records out
285212672 bytes transferred in 7.616 secs (37449142 bytes/sec)
=== mounting image via vnd vnd0 ===
/dev/rvnd0a: 272.0MB (557056 sectors) block size 8192, fragment size 1024
using 6 cylinder groups of 45.34MB, 5803 blks, 11264 inodes.
super-block backups (for fsck_ffs -b #) at:
32, 92880, 185728, 278576, 371424, 464272,
=== installing sets ===
base
etc
modules
=== performing customisations ===
=== making extra directories ===
=== customising /var/tmp ===
=== user customisations from files in custom ===
Filesystem 1K-blocks Used Avail %Cap Mounted on
/dev/vnd0a 269967 263958 -7489 102% /mnt
% ls -l -R custom
total 8
drwxr-xr-x 3 agc agc 512 Jan 18 03:58 etc
drwxr-xr-x 2 agc agc 512 Jan 18 04:00 iscsi

custom/etc:
total 8
drwxr-xr-x 2 agc agc 512 Jan 18 03:58 iscsi
-rw-r--r-- 1 agc agc 564 Jan 18 03:58 rc.conf

custom/etc/iscsi:
total 6
-r--r--r-- 1 agc agc 799 Jan 18 04:27 targets

custom/iscsi:
total 102484
-rw-r--r-- 1 agc agc 104857600 Jan 18 04:00 iscsi-target0
% sudo vnconfig vnd0 usermode.img
% sudo mount /dev/vnd0a /mnt
% cat /mnt/etc/iscsi/targets
# $NetBSD: targets,v 1.2 2006/02/16 19:30:57 agc Exp $
#
# Structure of this file:
#
# + an extent is a straight (offset, length) pair of a file or device
# it's the lowest common storage denominator
# at least one is needed
# + a device is made up of one or more extents or other devices
# devices can be added in a hierachical manner, to enhance resilience
# + in this example, no device definitions are necessary, as the target
# will just use a simple extent for persistent storage
# + a target is made up of 1 or more devices
# The code does not support RAID1 recovery at present

# Simple file showing 1 extent, mapped straight into 1 target

# extent file or device start length
extent0 /iscsi/iscsi-target0 0 100MB

# target flags storage netmask
target0 ro extent0 0.0.0.0/0
% ls -l /mnt/iscsi/iscsi-target0
-rw-r--r-- 1 agc agc 104857600 Jan 18 04:00 /mnt/iscsi/iscsi-target0
% diff custom/etc/iscsi/targets /mnt/etc/iscsi/targets
%

Add 3 scripts which are useful for making images for usermode, or any
embedded platform in particular.

+ mkimage will make an autosized image in a file (using vnd) from sets
the resulting image can be mounted read-only, with tmpfs used for volatile
files on top of read-only bases

+ mkpkgs will make an autosized image in a file (using vnd) from binary
packages

+ usermode is an easy way to invoke usermode, making sure that bridging
is set up properly, and that the host syscallemu module is loaded. This
script then runs usermode with two file images (as produced by mkimage
and mkpkgs)

branches: 1.1.4;
Add 3 scripts which are useful for making images for usermode, or any
embedded platform in particular.

+ mkimage will make an autosized image in a file (using vnd) from sets
the resulting image can be mounted read-only, with tmpfs used for volatile
files on top of read-only bases

+ mkpkgs will make an autosized image in a file (using vnd) from binary
packages

+ usermode is an easy way to invoke usermode, making sure that bridging
is set up properly, and that the host syscallemu module is loaded. This
script then runs usermode with two file images (as produced by mkimage
and mkpkgs)

branches: 1.1.4;
Add 3 scripts which are useful for making images for usermode, or any
embedded platform in particular.

+ mkimage will make an autosized image in a file (using vnd) from sets
the resulting image can be mounted read-only, with tmpfs used for volatile
files on top of read-only bases

+ mkpkgs will make an autosized image in a file (using vnd) from binary
packages

+ usermode is an easy way to invoke usermode, making sure that bridging
is set up properly, and that the host syscallemu module is loaded. This
script then runs usermode with two file images (as produced by mkimage
and mkpkgs)

branches: 1.2.4; 1.2.8;
merge more, and add an i386 image

Add preliminary support for creating x86 usb sticks with mkimage.
This is untested, and just a checkpoint of the work. Image looks
ok, but I have not tried if it boots.

Fix MKREPRO build (Jan-Benedict Glaw)

ec2_init: Add support for Oracle Cloud

Background dhcpcd if ec2_init is not enabled to speed up boot.

port-evbarm/56274: no network on ec2 arm64 9.99.85

Add -w to dhcpcd_flags when running on EC2, since we need to wait for the
network to come up before contacting the metadata service.

Create an "octeon.img.gz" bootable image for evbmips64-eb release builds.

Enable GPT support w/ hybrid MBR for Raspberry Pi compatibility.

Separate devicetree .dts -> .dtb building from kernel builds. They are now
part of a separate set, "dtb.tgz", and only built when MKDTB=yes. This
defaults to yes for earmv[67]* and aarch64, and no everywhere else.

RPi4 has a different root device, so use config.txt conditionals to load
an alternate cmdline txt file (cmdline-pi4.txt) on these boards.

Update to new RaspberryPi firware

commit 0c01dbefba45a08c47f8538d5a071a0fba6b7e83
Author: popcornmix <popcornmix@gmail.com>
Date: Wed Dec 11 15:30:08 2019 +0000

and include firmware for RPI4

Firmware has bee updated to support mainline linux kernels as described in
https://github.com/raspberrypi/linux/issues/3237

Fix whitespace in a comment

Add Amlogic dtb files to arm64.img

Add support for configuring Amazon.com EC2 SSH keys and hostnames. While
here, only set wscons=YES if a wsdisplay0 device is present.

Install ELF kernel to the FFS partition and boot it with bootaa64.efi
instead of loading kernel.img from the MSDOS partition with extlinux.conf

Move rk3399 dtb files to the correct location

branches: 1.2.2;
Install RK3328 .dtb files to the correct location

branches: 1.1.2;
Add initial support for ARMv8 (AARCH64) (by nisimura@ and ryo@)

- sys/arch/evbarm64 is gone and integrated into sys/arch/evbarm. (by skrll@)
- add support fdt. evbarm/conf/GENERIC64 fdt (bcm2837,sunxi,tegra) based generic 64bit kernel config. (by skrll@, jmcneill@)

Fix MKREPRO build (Jan-Benedict Glaw)

No need for ec2_init on arm64mbr

Create a separate Arm64 image that uses MBR partitioning. Amlogic SoCs
require the bootloader to be installed on SD cards at LBA1 which makes
them incompatible with GPT images.

Fix reproducible build (Jan-Benedict Glaw)

Install /boot/boot.cfg with EFI bootloaders.

On ARM ports, the EFI boot loader handles a boot.cfg file if it
exists, but one is not installed by default. In contrast, both the
i386 and amd64 ports do install boot.cfg by default. Bring these
ports into agreement by installing boot.cfg on ARM ports with EFI boot
loaders.

Update a few www.raspberrypi.com URLs

Move back to MBR based images for armv7 because Amlogic's bootloader (for
Amlogic S805 based boards) needs to be installed to sector 1, which
conflicts with the GPT header.

No need to rename netbsd-GENERIC.img to kernel7.img

Switch to GPT w/ hybrid MBR for armv7 images

Separate devicetree .dts -> .dtb building from kernel builds. They are now
part of a separate set, "dtb.tgz", and only built when MKDTB=yes. This
defaults to yes for earmv[67]* and aarch64, and no everywhere else.

Update to new RaspberryPi firware

commit 0c01dbefba45a08c47f8538d5a071a0fba6b7e83
Author: popcornmix <popcornmix@gmail.com>
Date: Wed Dec 11 15:30:08 2019 +0000

and include firmware for RPI4

Firmware has bee updated to support mainline linux kernels as described in
https://github.com/raspberrypi/linux/issues/3237

Move omap3 dtb files to /boot/dtb

Remove debug output, fix a progress message

Move a lot of *.dtb files to a dtb/ subdirectory on the FAT partition.
Mkimage (eroneously) creates a FAT16 partition (despite the configuration
asking for FAT32), and that has a root directory size limit.
Idea from Jared.

Switch to GENERIC kernels only.

branches: 1.35.2;
Fallback to non-EFI boot script if EFI bootloader is absent.

Note that we do not currently build EFI bootloader for earmv7hf-eb,
as it requires little-endian version of toolchain.

Switch the kernel7.img RaspberryPi kernel in the armv7 image to GENERIC
and thus reducing the number of kernels built in an evbarmv7 release.

Remove sun50i support from 32-bit GENERIC.

Install GENERIC ELF kernel as /netbsd for EFI booting

Install bootarm.efi to boot partition

No need to specify root and console bootargs on amlogic

Another kernel bites te dust.

Remove ODROID-C1 as it is now replaced by GENERIC. Thanks jmcneill.

Add support for configuring Amazon.com EC2 SSH keys and hostnames. While
here, only set wscons=YES if a wsdisplay0 device is present.

Make armv7.img boot on the DE0 Nano SoC. Requires u-boot from -current pkgsrc.

Use the GENERIC kernel where possible in armv7.img and remove the old
kernels from the bulid for SUNXI and TEGRA.

While I'm here disable *_INSTALL kernel builds for boards supported by
armv7.img

Whitespace

a bit of sorting

branches: 1.23.2; 1.23.4;
Add missing RPI FDT build goop, from nick.

On 64-bit boards, U-Boot expects to find .dtb files in an allwinner/ subdirectory

Remove legacy awin kernels from image

Be more precise about exactly what fails when something does.

Relying upon set -e to abort things is sort of OK (it is not
a recommended option to use in general - too many odd special cases),
but only if user can work out from the "build failed" what actually
went wrong.

Tested only on amd64 build (for this, i386 is the same) - if anyone
has problems on builds for other systems, please let me know. However
the changes affect only failure paths, the most likely problem would
be for a build to fail to halt on an error, and I hope I have avoided
that. There should be no difference at all to error-free builds.

Trailing whitespace

Don't install uEnv.txt for beagle; boot.scr handles this now.

Install a boot.scr on armv7.img that detects the running board and selects
the correct kernel and boot protocol. Should work on tegra, sunxi, and
am335x boards for now.

Create a config.txt for RPI and set 'enable_uart=1' and 'force_turbo=0'.
This config gives us a stable core_freq clock that is used to drive the
SD HOST and AUX UART devices.

More information on config.txt settings for the UART can be found here:

https://www.raspberrypi.org/documentation/configuration/uart.md

Add : to body of populate_sunxi to appease bash.

Test for kernel build directory before reading DTB list

Replace HUMMINGBIRD_A31 with SUNXI kernel on armv7.img and include .dtb
files for SUNXI and TEGRA kernels on the MSDOS partition.

branches: 1.12.4;
ODROID-C1: Now that the MBR partition table contains an entry for the BSD
partition, the disklabel no longer conflicts with the ODROID-C1 bootloader.
Root partition changes from ld0e to ld0a.

tegra: set loadaddr to 0x90000000 (the default kernel_addr_r overlaps
fdt_addr_r on TK1 + L4T 21.x), and set default root device to ld0a to
match new device probe order.

branches: 1.10.2; 1.10.4;
C&P of change in rpi.conf (maybe factor out)

By default, RPI firmware sets the max CPU frequency to 600MHz. This can be
overridden by setting arm_freq in config.txt, but the default freq at boot
is still 600MHz.

Add logic to rc.local to compare the current vs. max CPU frequency; if they
differ, set the target frequency to the maximum.

Remove JETSONTK1 and NYAN-BIG kernels, add a new unified kernel named TEGRA.

Cubieboard2 kernel is named netbsd-CUBIEBOARD.ub not netbsd-CUBIEBOARD2.ub

branches: 1.7.2;
install jetsontk1 boot script as boot.scr

Properly escape the $ when trying to pass the u-boot env var $kernel.
From Rin Okuyama.

Turn resize on

double quote police

add JETSONTK1 to armv7.img

Cleanup: make_filesystems is not used any longer, purge remains

Rename beagleboard.img to armv7.img. The new image includes the same
kernels as beagleboard.img plus support for Raspberry Pi 2, ODROID-C1,
Cubieboard2, Cubietruck, Hummingbird A31, and Banana Pi.

Rename beagleboard.img to armv7.img. The new image includes the same
kernels as beagleboard.img plus support for Raspberry Pi 2, ODROID-C1,
Cubieboard2, Cubietruck, Hummingbird A31, and Banana Pi.

escape $ in $board so this will boot on beaglebone too

branches: 1.17.4;
After discussions with cristos regarding the previous commit, he
convinced me it was sufficient to define GZIP_CMD only in
distrib/utils/embedded/mkimage and not again in the config files
beagleboard.conf, rpi.conf, and rpi_inst.conf similar to how the
other variables $src, $release, $mnt, ... are only defined there.

Fix build release of evbarm which had been failing in mkimage
beagleboard_smp with the elusive message:

"gzip: can't stat: gzip: No such file or directory"

This was due to GZIP=gzip being defined in mkimage; we now use:

GZIP_CMD=${TOOL_GZIP:-gzip} # ${GZIP} is special to gzip(1)

as distrib/sets/sets.subr already does in these files as well:

distrib/utils/embedded/mkimage
distrib/utils/embedded/conf/beagleboard.conf
distrib/utils/embedded/conf/rpi.conf
distrib/utils/embedded/conf/rpi_inst.conf

4th time's a charm (I hope). Make sure that we pick up the kernel path passed to
mkimage command line, and not the default.

Instead of trying to pass two kernels to beagleboard.conf, pass one and
derive the path to the second from the first.

Try to pick up the right location for BEAGLEBONE and BEAGLEBOARD kernels.

support beaglebone and beagleboard with the same image

branches: 1.11.4;
Use "gzip -dc" instead of "zcat" - some versions of zcat don't grok
gzip.

gzcat -> zcat

no objmachine by default.

remove trailing whitespace from EOF

make sure kernel is installed as netbsd.ub

keep beagleboard conf in sync with rpi

switch to makefs

Move mdnsd to beagleboard only. RPI doesn't like it at the moment.

branches: 1.3.2;
- Add support for xsets
- Factor out common evbarm code

set kerneldir= to the path used by build.sh

add mkimage config for generating beagleboard images

Recognize MKPOSTFIX=no in mkimage

Remove duplicated code, handle minwrites in _normal the same way we handle
it in _gpt. Initial patch from jmmv@.

branches: 1.42.2;
/etc/rc.d/cerctl_init: New script for certctl rehash in live images.

This is very limited -- it does not supplant postinstall to rehash
certificates on upgrade; it only runs certctl rehash if
/etc/openssl/certs is an empty directory, as you get in live images
not created with sysinst.

We could also have a more general-purpose way to run postinstall(8)
on first boot of an image, but that has a lot more moving parts to
think about, so let's start with this limited-scope low-risk
approach.

PR install/57629

XXX pullup-10

branches: 1.41.2;
Silence a warning when machdep.dmi.chassis-asset-tag is not present

ec2_init: Add support for Oracle Cloud

Disable kernfs on live images -- it is not required.

Fix fallout from mkimage rev 1.76.

For mkimage:

- Update "size" if auto-calculated.
- Use "dd bs=1" instead of non-portable "head -c".
- Some style nits.

For MD make_label() functions:

- Stop using "newsize" as image size in MB, use "size" instead.

Create an "octeon.img.gz" bootable image for evbmips64-eb release builds.

Allow config file to override hostname

Add GPT support to mkimage.

Remove the pretty much useless 128MB swap partition from the arm images.

branches: 1.33.2;
add a method to add user accounts or ssh keys to the embedded
(installable) images which may be run entirely headless and
have no current method to edit the installation without another
netbsd host to modify the root (FFS) partition.

creds_msdos reads the creds.txt file from the msdos boot
partition and provides 4 basic methods:

sshkeyfile <user> <path on msdos>
sshkey <user> <entry>
useraddhash <user> <passwd hash>
useradd <user> <passwd>

the first two create a user with ssh key(s), and the second
two create a user with a password. if the last method is used
and raw passwords are given, the creds.txt file will be
shredded and deleted by rm -P.


inspired by a request from a pine64 user. ok jmcneill@.

Fix variable escaping in dev_exists()

Add support for configuring Amazon.com EC2 SSH keys and hostnames. While
here, only set wscons=YES if a wsdisplay0 device is present.

Use special ROOT. prefix in fstab entries instead of assuming ld0

resize_disklabel: if disk and partition is not specified, use kern.root_device/kern.root_partition sysctls

branches: 1.28.2;
Increase reserved space at start of image from 4MB to 16MB to make room
for Rockchip bootloaders.

branches: 1.27.8;
minor cleanup in customize_evbarm():
keep handling of /etc/rc.conf in one place

branches: 1.26.2;
Slightly reduce the size of the MSDOS partition from 124M to 92M. With
the larger size, installing the ODROID-C1 bootloader causes the fs to
become unreadable. Not sure why..

Increase the size of the boot partition and remove the 'd' part from the
disklabel.

branches: 1.24.2; 1.24.4;
mount root partition with noatime

Align partitions to 64MB for the benefit of SD cards > 32GB

show resize_ffs progress bar when resizing root partition

Add support for auto-growing the root partition. Enable it for rpi.img.

Fix previous so that the new generated rc.conf is picked up and that
any new new mount points (e.g. /proc and /kern) are generated.

While here increate the fs size so that postfix can actually build
/etc/mail/aliases.db

add /var/shm tmpfs to fstab

PR/49596: mkimage: incorrect permissions on evbarm images

Use NetBSD.dist in our mtree specification to populate the top level
directories, etc.

For each generated file hand craft an mtree entry.

Remove specialdirs as NetBSD.dist creates /proc and /kern.

disable wapbl for evbarm images until stability issues are sorted out

turn on ntpd

branches: 1.15.4;
these changes produce a bootable system under qemu, but its root is trashed.

Add preliminary support for creating x86 usb sticks with mkimage.
This is untested, and just a checkpoint of the work. Image looks
ok, but I have not tried if it boots.

Fix /dev population of the evbarm images by using MAKEDEV -s and appending
to the mtree spec file passed to makefs.

Keep both c: and d: paritions so this can work with systems that have
RAW_PART as 2 or 3. Addresses PR/47688.

If minwrites is set:
- set noatime, nodevmtime for root
- use tmpfs for frequently written files
- use unix sockets instead of fifos for postfix
From:
http://rich-tbp.blogspot.com/2013/03/netbsd-on-rpi-minimizing-disk-writes.html

Disklabel need to use sectors, not bytes. Thanks to agc.

branches: 1.9.4;
Use HOST_SH to run scripts. Makes mkimage work on SunOS.

create /proc and /kern

set rc_configured=YES and a hostname

start with the original rc.conf

switch to makefs

Move mdnsd to beagleboard only. RPI doesn't like it at the moment.

Size for 512 byte sectors and base ffs partition size on totalsize minus
others.

branches: 1.2.2;
kernel.img is for rpi only

- Add support for xsets
- Factor out common evbarm code

Recognize MKPOSTFIX=no in mkimage

Remove duplicated code, handle minwrites in _normal the same way we handle
it in _gpt. Initial patch from jmmv@.

branches: 1.4.2;
/etc/rc.d/cerctl_init: New script for certctl rehash in live images.

This is very limited -- it does not supplant postinstall to rehash
certificates on upgrade; it only runs certctl rehash if
/etc/openssl/certs is an empty directory, as you get in live images
not created with sysinst.

We could also have a more general-purpose way to run postinstall(8)
on first boot of an image, but that has a lot more moving parts to
think about, so let's start with this limited-scope low-risk
approach.

PR install/57629

XXX pullup-10

branches: 1.3.2;
Disable kernfs on live images -- it is not required.

Fix fallout from mkimage rev 1.76.

For mkimage:

- Update "size" if auto-calculated.
- Use "dd bs=1" instead of non-portable "head -c".
- Some style nits.

For MD make_label() functions:

- Stop using "newsize" as image size in MB, use "size" instead.

Create an "octeon.img.gz" bootable image for evbmips64-eb release builds.

wii: Disable ntpd by default.

It takes up a lot of memory.

Recognize MKPOSTFIX=no in mkimage

Remove duplicated code, handle minwrites in _normal the same way we handle
it in _gpt. Initial patch from jmmv@.

branches: 1.1.2; 1.1.4;
evbppc: Add initial support for the Nintendo Wii

branches: 1.1.4; 1.1.8;
merge more, and add an i386 image

Create an "octeon.img.gz" bootable image for evbmips64-eb release builds.

Remove duplicated code, handle minwrites in _normal the same way we handle
it in _gpt. Initial patch from jmmv@.

branches: 1.2.2;
/etc/rc.d/cerctl_init: New script for certctl rehash in live images.

This is very limited -- it does not supplant postinstall to rehash
certificates on upgrade; it only runs certctl rehash if
/etc/openssl/certs is an empty directory, as you get in live images
not created with sysinst.

We could also have a more general-purpose way to run postinstall(8)
on first boot of an image, but that has a lot more moving parts to
think about, so let's start with this limited-scope low-risk
approach.

PR install/57629

XXX pullup-10

RISC-V support that works on QEMU with a single hart.

Thanks for Simon Burge for plic(4).

RISC-V support that works on QEMU with a single hart.

Thanks for Simon Burge for plic(4).

RISC-V support that works on QEMU with a single hart.

Thanks for Simon Burge for plic(4).

Update a few www.raspberrypi.com URLs

G/C unused.

Fix kernel= order, add pi0w conditional and comments

Conditionally boot kernel.img or kernel7.img depending on board

Separate devicetree .dts -> .dtb building from kernel builds. They are now
part of a separate set, "dtb.tgz", and only built when MKDTB=yes. This
defaults to yes for earmv[67]* and aarch64, and no everywhere else.

Update to new RaspberryPi firware

commit 0c01dbefba45a08c47f8538d5a071a0fba6b7e83
Author: popcornmix <popcornmix@gmail.com>
Date: Wed Dec 11 15:30:08 2019 +0000

and include firmware for RPI4

Firmware has bee updated to support mainline linux kernels as described in
https://github.com/raspberrypi/linux/issues/3237

branches: 1.34.4;
Fix spelling in error message

Add missing RPI FDT build goop, from nick.

Be more precise about exactly what fails when something does.

Relying upon set -e to abort things is sort of OK (it is not
a recommended option to use in general - too many odd special cases),
but only if user can work out from the "build failed" what actually
went wrong.

Tested only on amd64 build (for this, i386 is the same) - if anyone
has problems on builds for other systems, please let me know. However
the changes affect only failure paths, the most likely problem would
be for a build to fail to halt on an error, and I hope I have avoided
that. There should be no difference at all to error-free builds.

Trailing whitespace

Create a config.txt for RPI and set 'enable_uart=1' and 'force_turbo=0'.
This config gives us a stable core_freq clock that is used to drive the
SD HOST and AUX UART devices.

More information on config.txt settings for the UART can be found here:

https://www.raspberrypi.org/documentation/configuration/uart.md

Cleanup: make_filesystems is not used any longer, purge remains

Add support for auto-growing the root partition. Enable it for rpi.img.

Add RPI2 to kernels build for both earmv[67].

Use the earmv6 built kernels to create an image that can be used on both
RPI and RPI2

actually write cpufreq changes to ${mnt}/etc/rc.local instead of /tmp/a

By default, RPI firmware sets the max CPU frequency to 600MHz. This can be
overridden by setting arm_freq in config.txt, but the default freq at boot
is still 600MHz.

Add logic to rc.local to compare the current vs. max CPU frequency; if they
differ, set the target frequency to the maximum.

branches: 1.24.4;
Put console=fb back as autodetect is bunged up.

After discussions with cristos regarding the previous commit, he
convinced me it was sufficient to define GZIP_CMD only in
distrib/utils/embedded/mkimage and not again in the config files
beagleboard.conf, rpi.conf, and rpi_inst.conf similar to how the
other variables $src, $release, $mnt, ... are only defined there.

Fix build release of evbarm which had been failing in mkimage
beagleboard_smp with the elusive message:

"gzip: can't stat: gzip: No such file or directory"

This was due to GZIP=gzip being defined in mkimage; we now use:

GZIP_CMD=${TOOL_GZIP:-gzip} # ${GZIP} is special to gzip(1)

as distrib/sets/sets.subr already does in these files as well:

distrib/utils/embedded/mkimage
distrib/utils/embedded/conf/beagleboard.conf
distrib/utils/embedded/conf/rpi.conf
distrib/utils/embedded/conf/rpi_inst.conf

use console auto detection, ok nick

turn on devpubd

branches: 1.19.4;
Use "gzip -dc" instead of "zcat" - some versions of zcat don't grok
gzip.

gzcat -> zcat

dont use echo -n

handle renaming and uncompressing the kernel

add missing }

change to take the full kernel path

switch to makefs

cmdline needs to be a line! hi jak.

branches: 1.11.2;
Specify the root device in cmdline.txt, also, it's 'Pi', not 'PI'.

kernel.img is for rpi only

- Add support for xsets
- Factor out common evbarm code

now that we compute sizes correctly, reduce the extra to how much free space
we want.

make autosizing work (kind of)

first step at autosizing, still does not work.

append to rc.conf, don't overwrite

add the comp set

use variables for the fdisk command too

Autosize label

Adjust mkimage to handle RPI. From skrll@

Update a few www.raspberrypi.com URLs

Disable kernfs on live images -- it is not required.

rpi.img: Bump boot partition in a similar manner to other images.

Separate devicetree .dts -> .dtb building from kernel builds. They are now
part of a separate set, "dtb.tgz", and only built when MKDTB=yes. This
defaults to yes for earmv[67]* and aarch64, and no everywhere else.

Use fat16 as the partition is too small for fat32. from Harold Gutch

Use make_label_evbarm instead of home grown

upstream_kernel=1 is required with new firmware/kernel.

Update to new RaspberryPi firware

commit 0c01dbefba45a08c47f8538d5a071a0fba6b7e83
Author: popcornmix <popcornmix@gmail.com>
Date: Wed Dec 11 15:30:08 2019 +0000

and include firmware for RPI4

Firmware has bee updated to support mainline linux kernels as described in
https://github.com/raspberrypi/linux/issues/3237

branches: 1.11.4;
Build fix (really this time) from Harold Gutch

Be more precise about exactly what fails when something does.

Relying upon set -e to abort things is sort of OK (it is not
a recommended option to use in general - too many odd special cases),
but only if user can work out from the "build failed" what actually
went wrong.

Tested only on amd64 build (for this, i386 is the same) - if anyone
has problems on builds for other systems, please let me know. However
the changes affect only failure paths, the most likely problem would
be for a build to fail to halt on an error, and I hope I have avoided
that. There should be no difference at all to error-free builds.

Create a config.txt for RPI and set 'enable_uart=1' and 'force_turbo=0'.
This config gives us a stable core_freq clock that is used to drive the
SD HOST and AUX UART devices.

More information on config.txt settings for the UART can be found here:

https://www.raspberrypi.org/documentation/configuration/uart.md

More cleanup:
Instead of first overwriting the *_evbarm functions with own code,
and then calling them from the regular functions, directly put the
code into the regular functions.

No more functions from evbarm.conf are used now.
Some variables are still used!

Cleanup: make_filesystems is not used any longer, purge remains

PR/49596: mkimage: incorrect permissions on evbarm images

Use NetBSD.dist in our mtree specification to populate the top level
directories, etc.

For each generated file hand craft an mtree entry.

Remove specialdirs as NetBSD.dist creates /proc and /kern.

branches: 1.5.4; 1.5.6;
Include evbarm.conf as same as other conf files

evbarm.conf ensures ${mnt}/boot exists so that it avoids the error:
/home/builds/ab/HEAD/src/distrib/utils/embedded/mkimage: cannot create \
/tmp/mkimage.24444a/mnt/boot/cmdline.txt: directory nonexistent

After discussions with cristos regarding the previous commit, he
convinced me it was sufficient to define GZIP_CMD only in
distrib/utils/embedded/mkimage and not again in the config files
beagleboard.conf, rpi.conf, and rpi_inst.conf similar to how the
other variables $src, $release, $mnt, ... are only defined there.

Fix build release of evbarm which had been failing in mkimage
beagleboard_smp with the elusive message:

"gzip: can't stat: gzip: No such file or directory"

This was due to GZIP=gzip being defined in mkimage; we now use:

GZIP_CMD=${TOOL_GZIP:-gzip} # ${GZIP} is special to gzip(1)

as distrib/sets/sets.subr already does in these files as well:

distrib/utils/embedded/mkimage
distrib/utils/embedded/conf/beagleboard.conf
distrib/utils/embedded/conf/rpi.conf
distrib/utils/embedded/conf/rpi_inst.conf

branches: 1.2.6;
Fix an empty shell function here, per Greg Troxel. Apparently empty shell
functions are not 100% portable.

Add a new installer image for Raspberry PI to evbarm

Includes the following major changes:

1) Add new configmenu item to sysinst to allow creation of users at
post-install time.
2) Add an sshramdisk type to evbarm, which is a ramdisk with sshd
enabled, allowing users to install on a headless PI by sshing to the
device with the account "sysinst" password "netbsd" to begin the
installation. (Note, neither the account, nor the ssh setup are copied to
the finalized installation image)
3) Change mkimage to build a boot-only image, without the root filesystems.

Much of the sshramdisk code could be reused on other ports easily.

/etc/rc.d/cerctl_init: New script for certctl rehash in live images.

This is very limited -- it does not supplant postinstall to rehash
certificates on upgrade; it only runs certctl rehash if
/etc/openssl/certs is an empty directory, as you get in live images
not created with sysinst.

We could also have a more general-purpose way to run postinstall(8)
on first boot of an image, but that has a lot more moving parts to
think about, so let's start with this limited-scope low-risk
approach.

PR install/57629

XXX pullup-10

branches: 1.6.2;
Disable kernfs on live images -- it is not required.

Cleanup: make_filesystems is not used any longer, purge remains

PR/49596: mkimage: incorrect permissions on evbarm images

Use NetBSD.dist in our mtree specification to populate the top level
directories, etc.

For each generated file hand craft an mtree entry.

Remove specialdirs as NetBSD.dist creates /proc and /kern.

branches: 1.3.4; 1.3.12;
switch to makefs

branches: 1.2.2;
append to rc.conf, don't overwrite

Adjust mkimage to handle RPI. From skrll@

wii: Add standalone bootloader.

This adds "ppcboot.elf" to the Wii (evbppc) build, which can be placed on
the FAT partition of an SD card in the bootmii directory. The bootloader
doesn't have a user interface, it simply loads a kernel from the first
FFS partition it finds on the SD card.

Debug logs are printed to a USB Gecko device in either memory card slot
if present.

\\-__,------,___.
\\ __,---` NetBSD/evbppc Wii MINI boot
\\ `---,_. Revision 1.0 (Sun Nov 16 19:27:47 UTC 2025)
\\-,_____,.---`
\\
\\
\\

booting netbsd 7828064/-|+130520 [315696+305916]=0x82f350
[ 1.0000000] Copyright (c) 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003,
[ 1.0000000] 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013,
[ 1.0000000] 2014, 2015, 2016, 2017, 2018, 2019, 2020, 2021, 2022, 2023,
[ 1.0000000] 2024, 2025
[ 1.0000000] The NetBSD Foundation, Inc. All rights reserved.
[ 1.0000000] Copyright (c) 1982, 1986, 1989, 1991, 1993
[ 1.0000000] The Regents of the University of California. All rights reserved.

[ 1.0000000] NetBSD 11.99.3 (WII) #207: Sun Nov 16 14:28:23 AST 2025
[ 1.0000000] jmcneill@berserk:/home/jmcneill/netbsd/obj/sys/arch/evbppc/compile/WII
[ 1.0000000] total memory = 90112 KB
[ 1.0000000] avail memory = 77008 KB
[ 1.0000000] mainbus0 (root): Nintendo Wii
...

wii: Start bluetooth by default

branches: 1.4.2;
wii: update wii.img ffs parameters

Use 4K block/frags and FFSv2 for Wii images.

branches: 1.3.2; 1.3.4;
wii: Add 128x48 icon to SD card image

build fix: use dd with count=1 for compat with NetBSD dd(1)

evbppc: Add initial support for the Nintendo Wii

Recognize MKPOSTFIX=no in mkimage

Remove duplicated code, handle minwrites in _normal the same way we handle
it in _gpt. Initial patch from jmmv@.

branches: 1.11.2;
/etc/rc.d/cerctl_init: New script for certctl rehash in live images.

This is very limited -- it does not supplant postinstall to rehash
certificates on upgrade; it only runs certctl rehash if
/etc/openssl/certs is an empty directory, as you get in live images
not created with sysinst.

We could also have a more general-purpose way to run postinstall(8)
on first boot of an image, but that has a lot more moving parts to
think about, so let's start with this limited-scope low-risk
approach.

PR install/57629

XXX pullup-10

branches: 1.10.2;
Disable kernfs on live images -- it is not required.

Fix fallout from mkimage rev 1.76.

For mkimage:

- Update "size" if auto-calculated.
- Use "dd bs=1" instead of non-portable "head -c".
- Some style nits.

For MD make_label() functions:

- Stop using "newsize" as image size in MB, use "size" instead.

Be more precise about exactly what fails when something does.

Relying upon set -e to abort things is sort of OK (it is not
a recommended option to use in general - too many odd special cases),
but only if user can work out from the "build failed" what actually
went wrong.

Tested only on amd64 build (for this, i386 is the same) - if anyone
has problems on builds for other systems, please let me know. However
the changes affect only failure paths, the most likely problem would
be for a build to fail to halt on an error, and I hope I have avoided
that. There should be no difference at all to error-free builds.

Fix previous so that the new generated rc.conf is picked up and that
any new new mount points (e.g. /proc and /kern) are generated.

While here increate the fs size so that postfix can actually build
/etc/mail/aliases.db

PR/49596: mkimage: incorrect permissions on evbarm images

Use NetBSD.dist in our mtree specification to populate the top level
directories, etc.

For each generated file hand craft an mtree entry.

Remove specialdirs as NetBSD.dist creates /proc and /kern.

fix image building:
- round up the total size of the disk image
- fix fdisk offset for netbsd ffs images

branches: 1.4.4; 1.4.6; 1.4.8;
merge more, and add an i386 image

remove swap; these days x86 machines don't need it.

these changes produce a bootable system under qemu, but its root is trashed.

Add preliminary support for creating x86 usb sticks with mkimage.
This is untested, and just a checkpoint of the work. Image looks
ok, but I have not tried if it boots.

Install ELF kernel to the FFS partition and boot it with bootaa64.efi
instead of loading kernel.img from the MSDOS partition with extlinux.conf

branches: 1.1.2; 1.1.4;
Add initial support for ARMv8 (AARCH64) (by nisimura@ and ryo@)

- sys/arch/evbarm64 is gone and integrated into sys/arch/evbarm. (by skrll@)
- add support fdt. evbarm/conf/GENERIC64 fdt (bcm2837,sunxi,tegra) based generic 64bit kernel config. (by skrll@, jmcneill@)

Switch to GENERIC kernels only.

branches: 1.14.2;
Re-enable efiboot for armv7 (thanks skrll@)

Turn off efiboot until we figure out why it fails on some boards.

Switch to efiboot

Rename the DE0 Nano SoC dts file to its new linux/uboot name

Fix mmcpart for exynos SD card

Make armv7.img boot on the DE0 Nano SoC. Requires u-boot from -current pkgsrc.

Use the GENERIC kernel where possible in armv7.img and remove the old
kernels from the bulid for SUNXI and TEGRA.

While I'm here disable *_INSTALL kernel builds for boards supported by
armv7.img

Sort

Enable booting of netbsd-EXYNOS kernel

branches: 1.5.2; 1.5.4;
Enable automatic booting on tegra210

No need to specify console=fb for pinebook anymore as WSDISPLAY_MULTICONS will give us a framebuffer console

Use fb console by default on Pinebook

Fix test otherwise we try to load a fdtfile for boards which do not have use_fdt set.

Install a boot.scr on armv7.img that detects the running board and selects
the correct kernel and boot protocol. Should work on tegra, sunxi, and
am335x boards for now.

branches: 1.2.4;
No support for tegra210 in armv7 kernel

branches: 1.1.2;
Fallback to non-EFI boot script if EFI bootloader is absent.

Note that we do not currently build EFI bootloader for earmv7hf-eb,
as it requires little-endian version of toolchain.

Install /boot/boot.cfg with EFI bootloaders.

On ARM ports, the EFI boot loader handles a boot.cfg file if it
exists, but one is not installed by default. In contrast, both the
i386 and amd64 ports do install boot.cfg by default. Bring these
ports into agreement by installing boot.cfg on ARM ports with EFI boot
loaders.

properly quote arguments.

branches: 1.5.2;
remove clause 3 from all my licenses that aren't conflicting with
another copyright claim line. again. (i did this in 2008 and then
did not update all of my personal templates.)

branches: 1.4.4;
Fix typo

various cleanups inspired by reviews:
- clean up variable usage
- be explicit that $ssh_userkeys is global
- don't assig id output to a variable not used

explicitly strip carriage return so that non-unix text files work.

add a method to add user accounts or ssh keys to the embedded
(installable) images which may be run entirely headless and
have no current method to edit the installation without another
netbsd host to modify the root (FFS) partition.

creds_msdos reads the creds.txt file from the msdos boot
partition and provides 4 basic methods:

sshkeyfile <user> <path on msdos>
sshkey <user> <entry>
useraddhash <user> <passwd hash>
useradd <user> <passwd>

the first two create a user with ssh key(s), and the second
two create a user with a password. if the last method is used
and raw passwords are given, the creds.txt file will be
shredded and deleted by rm -P.


inspired by a request from a pine64 user. ok jmcneill@.

move creds_msdos.8 into the normal place and install it.

Use more markup, fix some typos.

add a method to add user accounts or ssh keys to the embedded
(installable) images which may be run entirely headless and
have no current method to edit the installation without another
netbsd host to modify the root (FFS) partition.

creds_msdos reads the creds.txt file from the msdos boot
partition and provides 4 basic methods:

sshkeyfile <user> <path on msdos>
sshkey <user> <entry>
useraddhash <user> <passwd hash>
useradd <user> <passwd>

the first two create a user with ssh key(s), and the second
two create a user with a password. if the last method is used
and raw passwords are given, the creds.txt file will be
shredded and deleted by rm -P.


inspired by a request from a pine64 user. ok jmcneill@.

ec2_init: Suppress error message for nonexistent sysctl.

We're querying to see whether the sysctl node is available and if so
what its text is, not interested in the error message.

XXX pullup-10

branches: 1.5.2;
ec2_init: Add support for Oracle Cloud

Extract just the random bits to feed to /dev/urandom.

This makes no difference in the randomness of the pool, but it improves
on the estimation (if any) of how many random bits were obtained.
Also make the ftp -q time out a bit longer since I got some time outs.

Add some OpenStack support.

I found that in the cloud I tried, by the time this script runs, there
is no default route in effect yet. That takes some 5 to 10 seconds
longer. So I added a retry loop, and to make that easier, changed the
order of queries. To make sure it doesn't wait ~forever for a
non-existent service I added the -q 1 option to ftp invocations.

I also added OpenStack-specific metadata which contains a different
random_seed of 512 bytes every time it is requested. See
https://github.com/openstack/nova/blob/master/nova/api/metadata/base.py#L355
It may not be trusted data but only in the strictest sense of the word.
The data can only be observed by people with access to the cloud's
overlay network for the particular VM.

AWS marketplace does not allow root ssh logins. Create an ec2-user account
and install the ssh key in that user's home directory instead.

branches: 1.1.2; 1.1.4;
Add support for configuring Amazon.com EC2 SSH keys and hostnames. While
here, only set wscons=YES if a wsdisplay0 device is present.

branches: 1.1.2;
wii: Add 128x48 icon to SD card image

branches: 1.2.2;
wii: Add support for passing boot options to the kernel.

evbppc: Add initial support for the Nintendo Wii

Can't use disklabel to find the real size of a disk since a fake disklabel
read from disk is trusted. Revert to use drvctl again.

Handle arbitrary raw partitions.
Fetch geometry from disklabel to be consistent.
Use disktab format instead of the normal human-readable output for parsing
the disklabel.
Avoid overflows in size computation.

resize_disklabel: if disk and partition is not specified, use kern.root_device/kern.root_partition sysctls

branches: 1.2.10; 1.2.12;
Grow the MBR partition table entry for the BSD partition before
growing the disklabel.

branches: 1.1.2; 1.1.4; 1.1.6;
Add support for auto-growing the root partition. Enable it for rpi.img.

Add support for hybrid MBR/GPT images.

Add GPT support to mkimage.