[BBB] nanobsd boot stuck

[Rodney W. Grimes]

> On Tue, Jan 01, 2019 at 12:47:26PM -0800, Rodney W. Grimes wrote:
> >> Hi,
> >> a freshly built nanobsd RELEASE-12.0-p1 for Beaglebone Black does not
> >> boot further than the line
> >> "Trying to mount root from ufs:/dev/mmcsd0s2a [ro]"
> >>
> >> I've tried to track this further down and the findings are
> >>
> >>
> >> 1. booting into single user mode works. When I boot into single user
> >> mode and exit from there, the BBB continues its boot into multiuser
> >> without any further problems. I'm able to login and have all my programs
> >> working.
> >>
> >> 2. setting "set -x" in /etc/rc to extend verbosity. After setting, i can
> >> see that the boot stops at line 62 of rc which says:
> >>
> >> dlv=`/sbin/sysctl -n vfs.nfs.diskless_valid 2> /dev/null`
> >
> >Hum, is /dev not getting mounted for some reason
> >causing an issue here.
> >Right before this line try adding:
> >#Lets see what file systems are here now
> >/bin/df
> >#Lets see if we have a /dev/null to use
> >/bin/ls -lag /dev/null
> >#Lets see what this command returns
> >/sbin/sysctl -n vfs.nfs.diskless_valid
> >>
> >> I altered this line to
> >>
> >> dlv=`/bin/echo "0"`
> >Uh, ah, I see it, this is NOT what happens when you run:
> >dlv=`/sbin/sysctl -n vfs.nfs.diskless_valid 2> /dev/null`
> >set | grep dlv
> >dlv=''
> >To get past try:
> >dlv=''
> >>
> >> but it still stops at that line. Using the builtin sh-echo instead of
> >> the binary one ( dlv=`echo "0"` ) the boot continues until:
> >
> >This is starting to sound like it is having issues accessing
> >the root device, like your sd card is toast.
> >>
> >> sh /etc/rc.initdiskless
> >>
> >Why did it wonder down the diskless path, you specifically
> >set dlv.. oh gone back and editied.. I see it, dlv=0, not same as dlv='';.
> >dlv gets null of vfs.nfs.dikless_valid does not exist as an oid:
> ># /sbin/sysctl -n vfs.nfs.diskless_valid
> >sysctl: unknown oid 'vfs.nfs.diskless_valid'
> >not 0 as you forced.
> 
> hmm, on all my machines (amd64 and armv7, 12.0-p1 and 11.2-STABLE) there 
> exists an oid named vfs.nfs.diskless_valid:
> 
> # sysctl -n vfs.nfs.diskless_valid
> 0
> 
> That was the reason why i've set dlv to 0 in the first place.

Argh, ok, I went down the wrong rabbit hole, teach me
for not doing a uname before I run stuff in an xterm
to see what the values are.  11.1 does not have this
oid.

> 
> >> After commenting this whole "diskless" section out, the boot continues
> >> by sourcing rc.subr and stops in the function load_rc_config() executing
> >> this line:
> >
> >>
> >> [ -r /etc/defaults/rc.conf ]
> >>
> >> 3. using different SD-Cards containing the nanobsd image makes no
> >> difference.  All cards show same behavior described above.
> >
> >sd card socket, bad sd code in the kernel image your using?
> 
> I've managed to organize another Beaglebone Black board, and on this one 
> the image does boot without issues. 
> It seems that the beaglebone i was developing on the whole time is 
> defect somehow!

I wonder if it is the sd card socket wore out from usage?
These are not really high duty cycle things, and why I
prefer to actually netboot these types of devices over
fooling with sd cards.   Though many have crappy network
performance, so that leads to other issues :-(.

> >> 4. regular FreeBSD Image (non-nanobsd) does boot correctly.
> >>
> >> I do not have a clue how to further debug this. Since the
> >> single-user-mode works, it seems to me that the installation as such is
> >> not entirely broken, but there is some hickup in the init procedure.
> >>
> >> Any ideas?
> >
> >Well a few corrections above might get you closer, since
> >you forced it down the diskless path, which is incorrect.
> 
> Thank you for your tips! 

Just glad its been found!

-- 
Rod Grimes                                                 rgrimes at freebsd.org