[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Installing on extended partition



Parodper, le lun. 23 août 2021 08:40:17 +0200, a ecrit:
> O 19/08/21 ás 20:34, Samuel Thibault escribiu:
> > > * The partition with the ISO would be hd0s6 (I did dd if=debian-hurd.iso
> > > of=/dev/sda6 on Linux), but if I try to write /dev/hd0s6 directly it fails,
> > > and trying to mount the partition manually fails with a «Input Output
> > > Error».
> > 
> > Ah! I hadn't understood that you were doing that.
> > 
> > You cannot do such thing with logical partitions: the dd call will wipe
> > the first sector which contains the logical partition chain cell. You
> > have to use a primary partition to be able to wipe out the first sector.
> > Alternately, since isofs doesn't actually use the first sectors, you can
> > skip one and seek one, to avoid wiping the logical partition chain cell.
> 
> Tried using «dd bs=512 skip=1 seek=1 conv=notrunc if=debian-hurd.iso
> of=/dev/sda8». It still needs MAKEDEV to mount it, and does not recognize
> the partition table.

I cannot reproduce the issue. In my test I could was able to just
provide the /dev/hd0s5 path and it'd just work.

> > > * Installer asks for drivers, choose to manually select the drive.
> > > * Then I change to TTY2, cd /dev and ./MAKEDEV hd0s6
> > > This last step was not necessary in QEMU, that is why I was asking.
> > 
> > That shouldn't be necessary at all, /dev is getting filled with device
> > translator entries during the boot process, see attached snapshot.
> > 
> I do get the same things on boot (see https://i.imgur.com/XGImhqK.jpg), but
> when I try to do «mount -t iso9660 /dev/hd08 /mnt», or even just «head
> /dev/hd0», I get a Input/Output error.

Ah, you mean that there really are some existing entries but they don't
work, and the entries created by MAKEDEV happen to seem to work?

About /dev/hd0 itself it's very surprising that it'd get Input/Output
errors since the entry that MAKEDEV create is strictly the same as the
existing one.

As for /dev/hd0s8, d-i indeed uses the parted layer to access the
partition, while MAKEDEV creates an entry that uses the kernel-level
driver for partitions. It is very surprising that the latter would
work better than the former, since only the former is maintained. When
you open your disk with parted from a Linux system, does it complain
somehow?

> PD: Is there any way to tell the translators to be more verbose? Something
> like 'echo OPTIONS=--verbose >> /etc/translator.conf'?

There is no such global thing since translators are allowed to be
implemented whatever the way they prefer. But you can start translators
by hand to make sure to get their output:

settrans -a /dev/hd0s8 /hurd/storeio -T typed part:8:device:hd0

Samuel


Reply to: