On Fri, May 05, 2017 at 03:52:22PM +0200, Andreas Wilhelm wrote:
> On 05/03/2017 02:13 PM, Sascha Hauer wrote:
> > On Tue, May 02, 2017 at 03:24:10PM +0200, Andreas Wilhelm wrote:
> >> Hello,
> >>
> >> I am using barebox state to exchange data between barebox and linux. From barebox I can access and change the state without any problems. If I try to access the state using the barebox-state util included with dt-utils (v2017.03.0), I get the following error:
> >>
> >>> # barebox-state -g myvar
> >>> of_get_devicepath: no 'label' property found in /soc/aips-bus@02000000/spba-bus@02000000/ecspi@02008000/m25p80@0
> >>> Cannot find backend path in /state
> >> This problem does not occur when using version 2016.08.0 of dt-utils. However, in this version I experienced an other problem. Even with backend-storage-type set to "circular" barebox attempted to storte data in "direct" mode:
> >>
> >>> WARNING, no stridesize given although we use a direct file
> >>> write. Starting in degraded mode
> >>> Failed to initialize desired amount of direct buckets, only 1 of 3
> >>> succeeded
> >> Furthermore changing the state from within barebox worked just fine, while changing it using dt-utils from within a booted linux didn't.
> >>
> >> The barebox state is stored in NOR-flash and configured in barebox as followed:
> >>
> >>> state: state {
> >>> magic = <0x27031977>;
> >>> compatible = "barebox,state";
> >>> backend-type = "raw";
> >>> backend = &flash, "partname:barebox-state";
> > Please use "backend = &barebox_state;" with a partition like this:
> >
> > barebox_state: partition@0 {
> > reg = <0x0 0x40000>;
> > };
> >
> >> Within the linux kernel I use the following configuration
> > Note that barebox will rewrite the state node in the device tree passed
> > to Linux to make sure that both are consistent. You do not need to add a
> > state node to the Linux device tree manually. Unfortunately the
> > rewriting of the device node does not work correctly if you use the
> > "partname:" binding above in barebox.
> >
> > Sascha
> >
> Thanks for your fast reply.
>
> Removing the state definition from the linux device tree and changing
> the backend definition stopped barebox and dt-utils from complaining
> about a bad state backend. However, now I am facing a strange error.
>
> I am able to read and change the barebox state from within barebox and
> changes are reflected when reading from within barebox and linux. When
> changing the state using dt-utils, the change is persistent between
> reboots, but cannot be seen from within barebox.
First of all, what storage device are you using? Is it a real flash or
some EEPROM or NVRAM?
>
> Example:
>
> // State myvar initialized with A by default
>
> state.myvar=B; state -s
> => Barebox: echo $state.myvar // myvar=B
> => Linux: barebox-state -g myvar // myvar=B
>
> barebox-state -s myvar=C
> => Barebox: echo $state.myvar // myvar=B
> => Linux: barebox-state -g myvar // myvar=C
This looks like a circular/noncircular mixup somewhere. Could you
hexdump the state partition (using the md command in barebox) to see if
the data is stored in circular or noncircular format? In circular format
you should see two copies of the same data directly after each other.
Also the output of barebox-state with -v might contain some useful
information.
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 |
_______________________________________________
barebox mailing list
barebox@xxxxxxxxxxxxxxxxxxx
http://lists.infradead.org/mailman/listinfo/barebox
