Re: udev auto-mount not doing fsck

[Denys Dmytriyenko <denis@denix.org>]

On Thu, Apr 10, 2014 at 05:47:02PM -0400, Brian Hutchinson wrote:
> On Thu, Apr 10, 2014 at 2:00 PM, Brian Hutchinson <b.hutchman@gmail.com>wrote:
> 
> > On Thu, Apr 10, 2014 at 12:56 PM, Denys Dmytriyenko <denis@denix.org>wrote:
> >
> >> Brian,
> >>
> >> Be careful, with your eUSB storage device cold-plugged (available on boot)
> >> you may run into a race condition, if you initially mount your rootfs in
> >> read-only mode and then re-mount in read-write mode during boot. One of
> >> the
> >> proposed solutions:
> >>
> >> http://thread.gmane.org/gmane.comp.handhelds.openembedded.core/50124
> >>
> >> --
> >> Denys
> >> --
> >> _______________________________________________
> >> Openembedded-devel mailing list
> >> Openembedded-devel@lists.openembedded.org
> >> http://lists.openembedded.org/mailman/listinfo/openembedded-devel
> >>
> >
> > At the moment the rootfs is rw in NOR.  In the future I plan for it to be
> > squashfs (ro) with a union overlay that is rw ubifs for persistent stuff.
> > This is to help protect the rootfs from getting corrupted or accidentally
> > hosed up during an update etc.
> >
> > So I could see where the race condition you mentioned could happen maybe
> > in that case but I'm not quite following (yet) how it could happen
> > currently in rw mode.  I guess you're saying there is a point where the
> > rootfs could be ro and so the udev rule has no mount point to hang the eUSB
> > on ... If I'm picking up what you're laying down ('if' being the key word
> > :) )
> >
> > This eUSB is intended to stay on the board at all times and not be hot
> > plugged ... but it is socketed (but board is buried in an enclosure) so it
> > could be replaced if needed by a technician.
> >
> > I'll study that patch you mentioned more when I get some time (in SCRUM
> > meeting at the time ... joy).
> >
> > Regards,
> >
> > Brian
> >
> >
> Hi Denys,
> 
> OK, I see what you are talking about now.  In my /etc/fstab I do have sixth
> column of root set to 1 so if I'm understanding things right
> /etc/rcS.d/S10checkroot.sh could be off checking root mounted ro while udev
> sees my eUSB and is in the process of doing RUN mount.sh and trying to
> mount /media/sda1 (in my case) before rootfs is remounted back rw.

Even w/o checking the rootfs, it's common to boot off of read-only FS and then 
remount it r/w slightly later - that's what checkroot.sh initscript is doing.

Unfortunately, udev starts slightly earlier than that and starts enumerating 
peripherals and firing its rules, including the auto-mount one, which can fail 
to create the necessary mount point. We had that happening intermittently, 
hence the proposed solution...


> Ah, I also see that /forcefsck only checks root ... for some reason I was
> thinking it would cause fsck to look at fstab and scan all filesystems in
> order of sixth column but that doesn't appear to be the case ... so that
> answers one of my questions.
> 
> I still don't know why my eUSB wasn't checked when I stubbed out udev
> automount script and instead added it to fstab with a 2 in the sixth column
> ... I thought that should of worked but it didn't.  It did in fact mount my
> eUSB but it did not check it.  The answer to that too is probably in one of
> these rcS scripts that I just haven't seen yet.
> 
> So the patch you mentioned is making more sense now.
> 
> Let me know if I'm tracking now or still off in the weeds.
> 
> Regards,
> 
> Brian
> -- 
> _______________________________________________
> Openembedded-devel mailing list
> Openembedded-devel@lists.openembedded.org
> http://lists.openembedded.org/mailman/listinfo/openembedded-devel