Re: preparing netbsd boot straps from gnu/linux

[pierre-philipp braun <pbraun%nethence.com@localhost>]

I switched to GPT it is a success with freebsd & UFS2, netbsd boots, fscks, and read-writes ffsv2 alright).  Booting is still done with legacy BIOS and GRUB2/knetbsd but now with the additional BIOS boot partition, an alternative to the Hybrid GPT trickery.

> | BAD SUPER BLOCK: VALUES IN SUPER BLOCK DISAGREE WITH THOSE IN FIRST ALTERNATE
> fsck_ext2fs compares primary and alternate superblock. Apparently there is
> a difference, unfortunately it's not shown.

Now trying the same but with ext2 from gnu/linux, I got that issue

	BAD SUPER BLOCK: VALUES IN SUPER BLOCK DISAGREE WITH THOSE IN FIRST ALTERNATE
	ioctl (GCINFO): Inappropriate ioctl for device
	/dev/rdk1: can't read disk label

So is that SUPER BLOCK issue truly reflecting what it says or is it simply about the fsck_ext2fs not finding a label for the device?


In case the message is true, here is the difference in two columns for a freshly built fs on gnu/linux.  I am not sure I should rather downgrade to -r0 or in the contrary try to make dir_index work without failing, but let's say that is the right way for now.

	mkfs.ext2 -O ^dir_index /dev/sda2

The primary superblock is on the left, and the first available secondary superblock on the right.

	http://tmp.nethence.com/diff.y.superblock.txt

For the record, this is how I reached the superblocks.  Checking block GROUP size

	tune2fs -l /dev/sda2 | egrep '^Block size' # 4096
	tune2fs -l /dev/sda2 | egrep '^Blocks per group' # 32768
	dumpe2fs /dev/sda2 | grep superblock | head

reading the ext2 primary and first backup superblocks (noticing `53 ef` at byte `56` https://wiki.osdev.org/Ext2)

	dd if=/dev/sda2 bs=1024 skip=1 count=1 2>/dev/null | hexdump -vC | less
	dd if=/dev/sda2 bs=1024 skip=$((4*32768)) count=1 2>/dev/null | hexdump -vC | less

	dd if=/dev/sda2 bs=1024 skip=1 count=1 > primary
	dd if=/dev/sda2 bs=1024 skip=$((4*32768)) count=1 > backup

	diff -y <(xxd primary) <(xxd backup)

Similarly here is how the superblock look like after I've deployed netbsd/amd64 8.0 on it manually

	http://tmp.nethence.com/diff.y.superblock2.txt

and this is the superblocks netbsd complains about at its first boot.

In the case the message is false and rather relates to labels

	disklabel dk1

gives indeed

	disklabel: ioctl DIOCGDINFO Inappropriate ioctl for device

> | /dev/rwd0e: NOT LABELED AS A EXT2 FILE SYSTEM (4.2BSD)
> It also verifies that the disklabel references it as an ext2 filesystem.

taking this into account although we're now talking about GPT/wedges.

when looking at `dmesg | grep ^dk` I see both wedges but with

	type: <unknown>

although it should be dk0 type `bios` and dk1 type `linux-data`

I've tried to fix this manually with `gpt type -T linux-data -i 2 wd0` and `dkctl wd0 makewedges` in vain.  The former returns fine but no change happens, while the latter only reports about removing/adding dk0, not dk1.

I did not re-check how the types are with the successful freebsd install PoC with GPT and UFS2.  fsck_ffs is not complaining at boot time.

-- 
Pierre-Philipp Braun
SNE Russia https://os3.su/