Re: Input/Output error booting some kernels on large disks

[Stephen Borrill <netbsd%precedence.co.uk@localhost>]

Further to this, this appears to be a BIOS problem when running in legacy 
mode (and machine is running latest available firmware). By adding 
debugging to /boot it was found that the bootloader was unable to access 
anything after 32GB. Therefore, my solution was to shrink the root 
partition from 40GB to 30GB (thanks resize_ffs!). With the netbsd-7 
bootloader this is fine. With -current, the bootloader has a long pause 
before it displays the boot menu and then another long pause before 
loading the kernel. During these pauses, the disk addresses suggest it is 
trying to read the 2nd GPT partition starting at 50GB. While these pauses 
are annoying, they don't stop it booting.

On Thu, 19 Mar 2020, Stephen Borrill wrote:

> I recently did a NetBSD 7 install on a pair of 14TB drives with GPT 
> partitioning. The disks are partitioned like this:
>
>        start         size  index  contents
>            0            1         PMBR (active)
>            1            1         Pri GPT header
>            2           32         Pri GPT table
>           34         2014         Unused
>         2048    104857600      1  GPT part - NetBSD RAIDFrame component
>                                 Type: raid
>                                 TypeID: 49f48daa-b10e-11dc-b99b-0019d1879648
>                                 GUID: c6775acf-756b-4cc9-8d57-35345ccff9e6
>                                 Size: 51200 M
>                                 Label: root1
>                                 Attributes: biosboot
>    104859648  27239905247      2  GPT part - NetBSD RAIDFrame component
>                                 Type: raid
>                                 TypeID: 49f48daa-b10e-11dc-b99b-0019d1879648
>                                 GUID: b2d43f4f-4689-44de-b9f0-27c8130fc395
>                                 Size: 12989 G
>                                 Label: data1
>                                 Attributes: None
>  27344764895           32         Sec GPT table
>  27344764927            1         Sec GPT header
>
> There's a RAID1 of the two 50GB GPT partitions (index 1) and I put a 
> disklabel on that to avoid GPT-on-RAID-on-GPT:
>
> # /dev/rraid0d:
> type: RAID
> disk: raid
> label: fictitious
> flags:
> bytes/sector: 512
> sectors/track: 128
> tracks/cylinder: 8
> sectors/cylinder: 1024
> cylinders: 81919
> total sectors: 104857472
> rpm: 3600
> interleave: 1
> trackskew: 0
> cylinderskew: 0
> headswitch: 0		# microseconds
> track-to-track seek: 0	# microseconds
> drivedata: 0
>
> 10 partitions:
> #        size    offset     fstype [fsize bsize cpg/sgs]
> a:  83886080         0     4.2BSD   4096 32768    90  # (Cyl.      0 - 
> 81919)
> b:  20969472  83886080       swap                     # (Cyl.  81920 - 
> 102397)
> d: 104857472         0     unused      0     0        # (Cyl.      0 - 
> 102399*)
>
> root is raid0a and is FFSv1. The system booted fine until I put on a newer 
> kernel (this is a non-uEFI install). I get open netbsd: Input/output error 
> from the bootloader when booting the new kernel. The original GENERIC kernel 
> still boots and a straight copy of the non-booting kernel also boots, so it's 
> not a kernel problem. I also tried /boot from NetBSD 9.0 and if anything this 
> was worse. With the 9.0 bootloader, there's an additional 10-second pause 
> between the primary bootstrap line and the /boot menu appearing.
>
> --
> Stephen