[BUG?] ubi.fm_autoconvert=1 on first boot

[Sascha Hauer]

Hi Richard,

On Fri, Mar 03, 2017 at 09:15:55AM +0100, Richard Weinberger wrote:
> Sascha,
> 
> Am 01.03.2017 um 10:42 schrieb Sascha Hauer:
> >> The patch works.
> > 
> > Hm, this patch works on mainline, but not on my 3.10 Frankenkernel. I
> > wonder if there's anything in mainline that makes sure that the wearlevel
> > worker has actually produced free PEBs before they are needed?
> 
> So it works every tine on mainline but never on 3.10?
> Just trying to make sure that it is not a timing issue.

It works on 3.10 when I add a mdelay(100) between starting the wearlevel
thread and autoresizing the volume, so indeed I assume it's a timing
issue.
Maybe I can provoke the failure on mainline with your patch applied by
adding a delay in the wear level thread.

> 
> In 3.10 Fastmap was very buggy and saw in later kernels many fixes,
> did you backport all of them? Sadly in the early days of Fastmap
> we did not mark patches for stable because it was experimental.
> ...which was a major error on our side.
> In doubt you can backport the whole UBI subsystem.
> This is rather easy because it has not much dependencies and does not change
> that fast.

My 3.10 has UBI/UBIFS patches backported from roughly 4.0, so
Frankenkernel is the right term to call it ;)
I remember backporting newer UBI patches at some point becomes difficult
because some newer kernel functions/macros are missing.

Sascha

-- 
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |