Re: mdconfig recently broken?

>Hi everybody,
>due to limited connectivity in August, I've been able to
>update my -CURRENT box only yesterday since the first days of August
>(iirc).
>
>I noticed that mdconfig -a -t vnode -f ${myfile} isn't working anymore
>when ${myfile} is on a read-only filesystem:
>
>sberta:/usr/local/freesbie-clone/uzip# mdconfig -a -t vnode -f usr.uzip
>md0
>sberta:/usr/local/freesbie-clone/uzip# mdconfig -d -u 0
>sberta:/usr/local/freesbie-clone/uzip# mount -fru /usr
>sberta:/usr/local/freesbie-clone/uzip# mdconfig -a -t vnode -f usr.uzip
>mdconfig: ioctl(/dev/mdctl): Read-only file system
>
>This makes impossible to use compressed filesystems on devices like
>cdrom or read-only diskless environment. Any idea? I'm using:
>
>FreeBSD sberta.saturnero.sat 7.0-CURRENT FreeBSD 7.0-CURRENT #2: Sat Aug 27 18:1
>5:39 CEST 2005     satu at sberta.saturnero.sat:/usr/obj/usr/src/sys/SBERTA  i386
>sberta:/usr/local/freesbie-clone/uzip#
>
>Bye and thanks in advance,
>Dario

Yes, this is the expected result. I changed this after it was noted that
it was possible to bypass schg flags through mdconfig. See kern/84635 for
more details. As a result the user needs to specify

	-o readonly

to mdconfig(8) when using read only files as backing stores.

phk and pjd both have expressed some interests in having me revert to a
graceful downgrade of access, which I am open to, provided we display
this fact to the user. I am just thinking if we should downgrade the
request in mdconfig instead of the kernel as it keeps the kernel code
more clean.

Here is a reference to the commit log, see revision 1.154

http://www.freebsd.org/cgi/cvsweb.cgi/src/sys/dev/md/md.c

Thoughts?

-- 
Christian S.J. Peron
csjp_at_FreeBSD.ORG
FreeBSD Committer