Two patches: Have vinum + dm_crypt at boot

[Rimvydas Jasinskas]

Hi,

I am not sure about etc/rc.d/cryptdisks change, these rc(8) bits can
be a bit tricky, someone else should take a look into it.
As for vinum(8) issue it should be fixed on master after bed5f6e.
http://tinyurl.com/j8n62yw

Thanks for reporting the issue.

Regards,
Rimvydas

On Fri, Nov 18, 2016 at 1:06 PM, Stefan Unterweger
<232.20711 at chiffre.aleturo.com> wrote:
> I am using a combination of dm_crypt and vinum to keep most of my data
> (sans root, which is on a different volume) in an encrypted RAID-1
> setup.  Specifically, physical slices are strung together with vinum,
> with a single LUKS volume (set up via ‘cryptsetup’) sitting on top.
>
> With the rc-scripts from base, this combination won’t boot on its own,
> since the ordering is slightly off:  ‘cryptdisks’ is run before
> ‘dumpon’, and ‘dumpon’ comes before ‘disks’.  Since ‘vinum’ provides
> ‘disks‘, rcorder will run it -after ‘cryptdisks‘, and the boot falls
> apart.
>
> This is easily solved:  By having ‘cryptdisks‘ explicitely come -after-
> ‘disks‘ (and not depending on ‘dumpon’), boot now first brings up vinum,
> and only -then- cryptdisks, and it -almost- boots:
>
> | ----------------------------- etc/rc.d/cryptdisks -----------------------------
> | index 3bc6232..d530437 100644
> | @@ -1,7 +1,8 @@
> |  #!/bin/sh
> |
> |  # PROVIDE: cryptdisks
> | -# BEFORE:  mountcritlocal localswap dumpon
> | +# BEFORE:  mountcritlocal localswap
> | +# REQUIRE: disks
> |
> |  $_rc_subr_loaded . /etc/rc.subr
>
> I’m still wondering what exactly I am breaking by not having ‘dumpon’
> ready when ‘cryptdisks’ comes up.  I have been running this patch for
> half a year and have not yet experienced any weirdness—but then again, I
> don’t really know what ‘dumpon’ exactly does…
>
>
>
> There‘s yet another glitch:  This early in the boot process, ‘/usr’ is
> not yet available, which makes perfect sense, since disks, volumes and
> stuff is still being brought up.  But ‘/sbin/vinum’ is dynamically
> linked and requires libraries which reside in ‘/usr/lib’:
>
> | Shared object ‘libdevstat.so.3‘ not found, required by ‘vinum’.
>
> I find it very weird that /sbin contains -any- dynamically linked
> binaries at all, especially binaries that require /usr already mounted,
> so this is perhaps an oversight.  The second patch patches the Makefile
> for ‘/sbin/vinum’ and forces static linking.
>
> With both patches, LUKS-on-top-of-Vinum boots like a charm:
>
> | ----------------------------- sbin/vinum/Makefile -----------------------------
> | index 9e6de2d..c427e0d 100644
> | @@ -7,7 +7,7 @@ WARNS?=       2
> |
> |  CFLAGS+=     -I${.CURDIR}/../../lib/libedit -I${.CURDIR}/../../sys
> |  DPADD=       ${LIBEDIT} ${LIBNCURSES} ${LIBDEVSTAT}
> | -LDADD=       -lprivate_edit -lprivate_ncurses -ldevstat
> | +LDADD=       -lprivate_edit -lprivate_ncurses -ldevstat -static
> |  LDFLAGS+=    ${PRIVATELIB_LDFLAGS}
> |
> |  .include <bsd.prog.mk>
>
>
>     Stefan
>
> --
> ▪ Die Internetbleibe.  Schick, magisch, leistungsstark.  https://internetbleibe.de/
> ▪ medoly media UG (haftungsbeschränkt) | Hausburgstraße 13, 10249 Berlin
> ▪ info at medolymedia.de | https://medolymedia.de/ | Tel. 030 609 826‒560 | Fax …‒569
> ▪ Geschäftsführer: Matthias Nothhaft | HRB 131198 (Amtsgericht Berlin-Charlottenburg), Sitz: Berlin, USt-ID: DE275221203