Disk management from Installation ISO

rvp

netbsdnoob 1) list the disks: i know i can do this with hw.disknames

That's the correct way to get all names. You can also use disklabel wd0 to list slices on a NetBSD MBR partition; or for GPT-partitioned disks, list wedges using dkctl wd0 listwedges.

netbsdnoob 2) check the size for each of the disks?

gpt -r header wd0                   # or, sd1, ld2, etc...

netbsdnoob 3) Check which partitions are on a disk and how much space they take?

gpt -r show wd0

netbsdnoob 4) Format disks/partitions

After you add GPT partitions using gpt:

dkctl wd0 listwedges
newfs -O2 /dev/rdk10              # use correct wedge name
newfs -O2 NAME=GPT_label          # OR, you can also use a GPT label instead

netbsdnoob

rvp what is the naming convention here?

I see on my system wd0 and dk1 dk2. How do i interpret these?

Also is there a more human-friendly way of showing size with gpt -r show wd0 when it comes to the size?
I mean showing e.g., 8GB instead of 16156623 or something like that.?

netbsdnoob

rvp @JuvenalUrbino ok now let's say i booted from a live image,

and want to prepare a disk for installing netbsd without sysinst, what are the minimum number of partitions i need if i am using gpt, it would be interesting to know with and without UEFI.
So gpt + bios and gpt + UEFI

rvp

netbsdnoob I see on my system wd0 and dk1 dk2. How do i interpret these?

wd0 is the whole (S)ATA disk. SCSI disks would be sd0. NVME disks ld0 (I think).
dk1 and dk2 are wedges the system creates from the partitions on a GPT-partitioned disk. The installation section in the Guide has more details.

netbsdnoob Also is there a more human-friendly way of showing size with gpt -r show wd0 when it comes to the size?
I mean showing e.g., 8GB instead of 16156623 or something like that.?

Well, there's a reason for that sector display: it's very precise--you can pretty easily see, to take an example, that there are no overlaps in the partitions.

Should be pretty easy to write a "humanizing" front-end like this

JuvenalUrbino

netbsdnoob I see on my system wd0 and dk1 dk2. How do i interpret these?

wd* is for ATA drives
sd* is for SCSI and USB drives
ld* is for logical disks (this includes SDs and PCIe drives)
md* is for memory disks
dk* is a form of hardware abstraction on NetBSD, grouping all the above under a disk-like interface, the so called wedge. Partitions inside physical disks are configured as dk wedges by the kernel as soon as the disk is attached. The dk(4) driver is conceptually similar to FreeBSD's GEOM.

JuvenalUrbino

rvp the system creates from the partitions on a GPT-partitioned disk.

It can work also on other partition layout forms (MBR, BSD label, Amiga RDB). But only GPT and APM are configured in the GENERIC kernel:

~ $ config -x /netbsd | grep DKWEDGE_METHOD
options         DKWEDGE_METHOD_GPT      # Supports GPT partitions as wedges
#options        DKWEDGE_METHOD_BSDLABEL # Support disklabel entries as wedges
#options        DKWEDGE_METHOD_MBR      # Support MBR partitions as wedges
options         DKWEDGE_METHOD_APPLE    # Support Apple partitions as wedges
#options        DKWEDGE_METHOD_RDB      # Support RDB partitions as wedges

rvp

JuvenalUrbino It can work also on other partition layout forms (MBR, BSD label, Amiga RDB). But only GPT and APM are configured in the GENERIC kernel:

Yah, and I've never really understood why DKWEDGE_METHOD_APPLE is enabled in an x86 GENERIC but DKWEDGE_METHOD_MBR isn't.

rvp

netbsdnoob o gpt + bios and gpt + UEFI

You mean BIOS + MBR and UEFI + GPT? (any other combination is pretty uncommon)

For UEFI + GPT, on a new disk, you need at least 3 partitions:

An EFI System Partition (MS-DOS/VFAT/...) for the EFI bootloader files
NetBSD EFI partition for /
NetBSD swap partn.

For BIOS + MBR, you need 1 primary NetBSD-type partition. You then divide this NetBSD partition using disklabel to create a (for /) and b (for swap) "slices".

All this should be there in the Guide.

JuvenalUrbino

rvp any other combination is pretty uncommon

I'd argue that GPT on BIOS on NetBSD is rather common; works fine, is integrated in sysinst, and has been supported for a very long time.

     gpt biosboot [-A] [-c bootcode] [-b startsec] [-i index] [-L label]
             The biosboot command allows the user to configure the partition
             that contains the primary bootstrap program, used during boot(8).

         The -A options sets the PMBR partition active.  This should not
         normally be necessary, but some firmware might require it.  If -A
         is omitted, the active flag will be cleared from the PMBR label.

         The -c option allows the user to specify the filename from which
         gpt should read the bootcode.  The default is to read from
         /usr/mdec/gptmbr.bin.

         The partition that should contain the primary bootstrap code,
         (similar to that installed via installboot(8)) is selected using
         the -i, -L and -b options.  One of these three options is
         required.  The -i option selects the partition given by the
         index.  The -L option selects the partition by label.  If there
         are multiple partitions with the same label, the first one found
         will be used.  The -b option selects the partition starting at
         block startsec.

rvp

JuvenalUrbino I'd argue that GPT on BIOS on NetBSD is rather common; works fine, is integrated in sysinst, and has been supported for a very long time.

There is also a mini-example at the end in that gpt(8) manpage for BIOS + GPT, but I've never tried it out. Nor have I tried persuading sysinst to do it for me (it's a temperamental beast at the best of times, and don't want it trashing my partitions).

JuvenalUrbino

rvp There is also a mini-example at the end in that gpt(8) manpage for BIOS + GPT, but I've never tried it out. Nor have I tried persuading sysinst to do it for me (it's a temperamental beast at the best of times, and don't want it trashing my partitions).

I've done it since NetBSD-7 days or so, and for what it's worth, in my case it just worked 😉.