Re: Remounting as read-only

[jgw%freeshell.org@localhost]

Stephen Borrill <netbsd%precedence.co.uk@localhost> wrote:

> On Fri, 14 May 2010, Steven Bellovin wrote:
> >>> That updates it to rw, which is not what you want. There was another 
> >>> person
> >>> asking for this in NetBSD 5.0. Turns out it can't sync meta data upon
> >>> remount read only: http://gnats.netbsd.org/30525
> >>
> >> I'm not sure what you mean by "it", but I should point out that the bug
> >> in question -- kernel doesn't sync metadata on read-only remount -- is
> >> extremely old, and *did* cause disk corruption after r/o remount set
> >> the filesystem "clean" flag, thus corruption fsck would never find and
> >> which could potentially eat your disk over time.
> >>
> >> It's *always* been broken.  Not just in NetBSD, probably in 4.4BSD
> >> and prior kernels as well.  It is unquestionably broken at least in older
> >> releases of the other BSDs as well. We have opted for safety over a false
> >> sense of security.
> >
> > Yup, understood.  I just want it to work, and work properly; it's a useful 
> > feature for embedded systems:
> >
> >     # uptime
> >     11:21AM up 593 days, 14:32, 0 users, load averages: 0.06, 0.08, 0.08
> >
> > With, probably, half a dozen configuration changes during that time.
>
> *nod* That's exactly how I use it too (in thousands of machines in 
> production daily with no problems). I suppose it might be argued that you 
> could use WAPBL now.

I'm curious if there are any "best practices" for those wishing to install
NetBSD in its current form (ie. WAPBL, ptyfs, kernel modules, ??) on
compact flash (and similar media).  Most of the guides seem to be several
years old now and, to a non-developer like myself, it's unclear what tweaks
should be made.  In the case of WAPBL (journaling) I'd think it ought to
be turned off to prevent excessive disc writes, no?

Regards,
Jeff