Hello, On Tue, Jan 22, 2019 at 09:53:37AM +0100, Uwe Kleine-König wrote: > Hello, > > On Tue, Jan 22, 2019 at 09:10:40AM +0100, Scherer, Thorsten wrote: > > i spent some time trying to understand the usage of imd_used etc. > > > > In another reply I stated: > > > > 'I misread the documenation on IMD as "of_compatible and model will be > > figured out and put into the proper location by barebox".' > > That is wrong. The current code doesn't ensure that and I don't see a > way how this could be done comfortably without per-board changes. > > > I am not so sure, if that is really the case. > > > > As far as i understand things: > > > > 1. of_compatible and model are (should be) automagically added to imd from the > > entries in the dts. > > > > git show 97e81f2d7 | grep -A 1 'For.*model.*of_compatible' > > cat scripts/gen-dtb-s | grep -A 23 -B 3 'barebox_imd_0' > > The imd tags for compatible and model are generated, but not > automatically included with this change. This is AFAICT the current > state. > > The problem is that the needed information is included in the "barebox" > image (i.e. the file with the name "barebox" that is build in the top > build dir) but then there is an additional step that extracts "all > needed" parts from that image to create a binary that then is able to be > started on your target. So if "barebox" has three entry points for > different machines, there are three images generated and each only has a > single entry point. Unused functions and data structures are stripped. > The linker detects that the imd section that contains the compatible and > model is not used and so this goes away in all three generated binaries. > Thank you for the explanation. Now things start to clear up a bit. > > 2. the IMD_USED macro puts arbitrary tags in the metadata. Which may > > also be a tag named model. > > The IMD_USED claims a certain data structure is used such that the > linker doesn't throw it away. > Ok. > > Now I wonder, if the lack of the patch just covers up an issue at another > > place or if there is something (in the context of multi image support?) which > > I do not understand. > > Did you look at the patch I sent you? > I did. But not closely enough. On the invokation of imd, model and of_compatible were not displayed. Somehow I did not get, that your patch was made to apply for the imx25 boards and not for the imx6 board (the one which I was checking). After adjusting the lowlevel.c file of the imx6 board, everything went fine. > > Maybe there is something messed up with the entries in > > the device tree? > > > > https://www.spinics.net/lists/arm-kernel/msg699453.html > > This is unrelated, this is a patch for Linux to add a device tree. Right. It didn't make sense to add this. > > Best regards > Uwe > Again, thank you for feedback. Best regards Thorsten > -- > Pengutronix e.K. | Uwe Kleine-König | > Industrial Linux Solutions | http://www.pengutronix.de/ | -- Mit freundlichen Grüßen Thorsten K. Scherer Diplom-Ingenieur (FH) Kälte- und Gebäudeleittechnik Datentechnik Eckelmann AG Berliner Str. 161, 65205 Wiesbaden Telefon +49 611 7103-329 Fax +49 611 7103-133 thorsten.scherer@xxxxxxxxxxxx https://www.eckelmann.de Eckelmann Group - Source of inspiration _______________________________________________ barebox mailing list barebox@xxxxxxxxxxxxxxxxxxx http://lists.infradead.org/mailman/listinfo/barebox
