Re: Corruped NAND booting for all devices

[Steve deRosier <derosier@xxxxxxxxx>]

On Wed, Feb 5, 2020 at 4:49 PM JH <jupiter.hce@xxxxxxxxx> wrote:
>
> Hi Jef,
>
> On 2/5/20, Jef Driesen <jef.driesen@xxxxxxx> wrote:
> >> [    6.782640] UBIFS (ubi0:0): UBIFS: mounted UBI device 0, volume 0,
> >> name "rootfs_data", R/O mode
> >>
> >> ...
> >>
> >> [FAILED] Failed to mount /var/volatile.
> >> See 'systemctl status var-volatile.mount' for details.
> >> [DEPEND] Dependency failed for Bind mount volatile /var/cache.
> >> [DEPEND] Dependency failed for Bind mount volatile /srv.
> >> [DEPEND] Dependency failed for Bind mount volatile /var/spool.
> >> [DEPEND] Dependency failed for Bind mount volatile /var/lib.
> >
> > At first sight, it looks you have a read-only ubifs filesystem, with an
> > overlay filesystem backed by another read-write ubifs filesystem? And
> > that read-write filesystem fails to mount after a power failure?
>
> I did notice the R/O mode, it was never in previous hardware revision,
> the MTD partition for UBI is always RW, I did not know where is that
> from, since the MTD partitions and installation was performed by our
> hardware contractor, I don't believe it could be my Yocto image to
> change it, but correct me, I could be wrong.
>

When UBIFS encounters an error, it switches the partition to R/O.
That's not a hardware thing necessarily.

Also, "MTD partitions and installation was performed by our hardware
contractor"?!?!  You've got to figure out the setup. Are you saying
that they're 100% responsible for everything below the level of
userspace? Or are you responsible for kernel, u-boot, etc... all the
software?  If so, this is a software thing and you've got to own it.
If they are truly responsible for everything lower than the userspace
software, it's time you stop wasting your time on it and get them to
debug it, this is way out of your realm.

> > In that case, this sounds very similar to the problem I reported last week:
> >
> > http://lists.infradead.org/pipermail/linux-mtd/2020-January/093542.html
>
> Interesting, good to know there is an issue here,
>

This is not necessarily your problem. Read and understand it, and run
some tests to check. But, t isn't likely the same thing.

> I guess the question here, firstly why we had the RO ubifs filesystem
> backed by an overlay RW ubifs filesystem in the first place, secondly
> is it an MTD bug or is it our fault to set up wrong RO and overlay RW
> ubifs?
>

Do you know you have this setup or not?  Based on what you're saying,
I don't think you have any idea. Boot a working device and check to
see if are running an overlayfs.

- Steve

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/