Re: xbd and large sectors

[Christian Kujau <lists%nerdbynature.de@localhost>]

On Fri, 21 Jul 2023, Manuel Bouyer wrote:
> In this case you can just use
> ftp://asim.lip6.fr/outgoing/bouyer/rel/HEAD/amd64/
> or wait for a recent-enough build to show up on
> http://nyftp.netbsd.org/pub/NetBSD-daily/HEAD/

Thanks, that's helpful of course. So, my options were either to wait until 
the builds show up, or checkout the latest trunk and build it myself. So, 
I did both.

After a few tries (and adding more diskspace for the build), a strange 
build error[0] and some hours later, a shiny new[1] install kernel was 
generated, yay.

Booting this in PV mode, these weird disk errors were back, but at least 
the disk were now recognized with the correct size (8G):

 [   1.1300569] xbd0: 8192 MB, 512 bytes/sect x 16777216 sectors
 [   1.1300569] xbd0: backend features 0xc<PERSISTENT,INDIRECT>
 [   1.1300569] xbd0d: error reading fsbn 0 (xbd0 bn 0; cn 0 tn 0 sn 0)
 [   1.1300569] xbd0: dos partition I/O error
 [   1.1400549] xbd0d: error reading fsbn 0 of 0-31 (xbd0 bn 0; cn 0 tn 0 sn 0)
 [   1.1400549] xbd0d: error reading fsbn 1 (xbd0 bn 1; cn 0 tn 0 sn 1)
 [   1.1700597] xbd1: 16384 MB, 512 bytes/sect x 33554432 sectors
 [   1.1700597] xbd1: backend features 0xc<PERSISTENT,INDIRECT>
 [   1.1700597] xbd1d: error reading fsbn 0 (xbd1 bn 0; cn 0 tn 0 sn 0)
 [   1.1700597] xbd1: dos partition I/O error

Hope! But, the *partition* step in sysinst failed again:

    Status: Command failed
    Command: disklabel -w -r -f /tmp/disklabel.6598 xbd0 'xbd0' 'default label'
    Hit enter to continue
    disklabel: ioctl DIOCWDINFO: Input/output error

For the record, here's the "disklabel xbd0" output again when booting 9.3 
in PV mode, notice how the sector and cylinder count are way off and it 
sees the disk as 64G in size:

--------- NetBSD 9.3 -------------------------
# disklabel xbd0
[  21.3900372] xbd0d: error reading fsbn 0 (xbd0 bn 0; cn 0 tn 0 sn 0)
# /dev/rxbd0:
type: ESDI
disk: xbd0
label: default label
flags:
bytes/sector: 512
sectors/track: 2048
tracks/cylinder: 1
sectors/cylinder: 2048
cylinders: 65536
total sectors: 134217728
rpm: 3600
interleave: 1
trackskew: 0
cylinderskew: 0
headswitch: 0           # microseconds
track-to-track seek: 0  # microseconds
drivedata: 0 

4 partitions:
#        size  offset  fstype [fsize bsize cpg/sgs]
 a: 134217728      0   4.2BSD      0     0     0  # (Cyl. 0 -  65535)
 d: 134217728      0   4.2BSD      0     0     0  # (Cyl. 0 -  65535)
disklabel: boot block size 0
disklabel: super block size 0
disklabel: partitions a and d overlap
----------------------------------------------

And here it is when booted with HEAD[1] in PV mode:

--------- NetBSD trunk -----------------------
# disklabel xbd0
[ 1042.4000859] xbd0d: error reading fsbn 0 (xbd0 bn 0; cn 0 tn 0 sn 0)
[ 1042.4000859] xbd0: dos partition I/O error
# /dev/rxbd0:
type: ESDI
disk: xbd0
label: default label
flags:
bytes/sector: 512
sectors/track: 2048
tracks/cylinder: 1
sectors/cylinder: 2048
cylinders: 8192
total sectors: 16777216
rpm: 3600
interleave: 1
trackskew: 0
cylinderskew: 0
headswitch: 0           # microseconds
track-to-track seek: 0  # microseconds
drivedata: 0 

4 partitions:
#        size  offset  fstype [fsize bsize cpg/sgs]
 a:  16777216    0     4.2BSD      0     0     0  # (Cyl. 0 - 8191)
 d:  16777216    0     4.2BSD      0     0     0  # (Cyl. 0 - 8191)
disklabel: boot block size 0
disklabel: super block size 0
disklabel: partitions a and d overlap
----------------------------------------------

By the time I built the kernel myself, the latest kernels showed up on the 
mirrors and just to be sure I booted 202307220710Z in PV mode and the same 
happens: the disk size is calculated correctly, but it still errors out 
with the messages above.

So, thanks again for trying to fix this. With my build cache still warm, 
the next build could be faster and I can respond sooner with test results. 
So, if you have more patches to test, please share :-)

Christian.

[0] ERROR: nbctfmerge: Input file synaptics.o was partially built from C sources, but no CTF data was present
[1] changeset 368775:ed7b96ef7fc5 "Also use XEN_BSIZE when computing the number of bytes for format_bytes()"
-- 
BOFH excuse #147:

Party-bug in the Aloha protocol.