Re: CVS commit: src/sys/kern

[Izumi Tsutsui <tsutsui%ceres.dti.ne.jp@localhost>]

christos@ wrote:

> On Jul 14,  9:20am, tsutsui%ceres.dti.ne.jp@localhost (Izumi Tsutsui) wrote:
> -- Subject: Re: CVS commit: src/sys/kern
> 
> | christos@ wrote:
> | 
> | > Module Name:      src
> | > Committed By:     christos
> | > Date:             Fri Jul 13 16:15:49 UTC 2012
> | > 
> | > Modified Files:
> | >   src/sys/kern: subr_disk_mbr.c
> | > 
> | > Log Message:
> | > revert previous; the problem was off by one in the bios device comparison
> | > in x86_autoconf.c
> | 
> | Sorry, the off by one is my fault in rev 1.35, but it has been there
> | before 5.0 and fixing x86_autoconf.c doesn't make "root on cd0a" work
> | without this workaround at least on my Optiplex 760.
> | 
> | I think the problem in subr_disk_mbr.c is that it makes
> | /dev/cd0a unconfigured if any unexpected error occurs during
> | reading media because kernel recognizes boot device as cd0 properly:
> | http://mail-index.NetBSD.org/current-users/2012/06/13/msg020407.html
> 
> Ok, so do we need to put this back for now,

I think so.

> or should someone determine
> what the unexpected media error is?

It's just a guess because bootloader also fails to load boot.cfg
in installcd. (I can test if someone provides debug code though)

Anyway I don't think it's a good idea to fake a disklabel
by reading data structures in target media in subr_disk_mbr.c.
It looks inconsistent with cdgetdefaultlabel() in sys/dev/scsipi/cd.c
and I don't see reason why UDF is in RAW_PART (not partition a:)
if we can detect FS type by reading media.

Did we have any discussion about the design?
If not, we should put the workaround (that disables override)
for 6.0.

---
Izumi Tsutsui