Re: MEMORY_RBFLAGS in options(4)

[Jean-Yves Migeon <jean-yves.migeon%espci.fr@localhost>]

David Brownlee wrote:
> On Wed, 17 Dec 2008, Jean-Yves Migeon wrote:
> > > > For this reason, I would like to know if it is possible to:
> > > > - rename the MEMORY_RBFLAGS option to MEMORY_DISK_RBFLAGS (to be in 
> > > > sync with the rest of the MEMORY_DISK_* macros)
> > > > - make adjustments inside src/ to reflect this
> > > > - document it inside options(4) so it is easier to use it
> > > > - declare "MEMORY_RBFLAGS" as obsolete (perhaps in conf/files?)
>>>>
> > > > [snip]
>>>>
> > > That seems very reasonable. I would go as far as to switch the 
> > > default (and update in tree md users to define MEMORY_DISK_RBFLAGS 
> > > for single user mode...
> >
> > You mean switching the default to, say, RB_AUTOBOOT instead of RB_SINGLE?
>
>     Yes, exactly.

So, here are the diffs that modify the MEMORY_DISK_RBFLAGS behaviour, 
and set RB_AUTOBOOT as default (was RB_SINGLE previously):

http://www.netbsd.org/~jym/RBFLAGS.diff
http://www.netbsd.org/~jym/options.4.diff

For config(5) files expecting MEMORY_DISK_RBFLAGS to be RB_SINGLE, I set 
it manually inside it. I did it with following criterias:

- if MEMORY_RBFLAGS was set explicitly: keep it.

(3 files do so: evbarm/conf/ARMADILLO210, evbarm/conf/GUMSTIX and 
ia64/conf/GENERIC.SKI)

- if MEMORY_DISK_HOOKS and MEMORY_DISK_IS_ROOT are set: set 
MEMORY_DISK_RBFLAGS to RB_SINGLE (to match old behaviour)

(71 configs, mostly ramdisks and install kernels)

- if MEMORY_DISK_HOOKS is set alone: do not touch. If I understand the 
code right, boothowto can only be overriden when MEMORY_DISK_IS_ROOT is 
set, or when md_root_setconf() is called when MEMORY_DISK_DYNAMIC is set 
(and with root on md0). Since neither are used in this case, I consider 
that boothowto cannot be altered by RBFLAGS.

(this is the case for atari)

- if MEMORY_DISK_HOOKS and MEMORY_DISK_DYNAMIC are set: investigate. 
This happens only for GENERIC i386 and amd64, so I decided _not_ to set 
RBFLAGS to RB_SINGLE as it could be misleading in the config(5) file 
(although this does not seem to affect GENERIC behaviour, as root is not 
on md0).

Thanks for the comments and/or reviews during this snowy and cheerful 
context :)

Cheers,

--
Jean-Yves Migeon
jean-yves.migeon%espci.fr@localhost