On Mon, May 31, 2021 at 09:12:33AM +0200, Ahmad Fatoum wrote: > The request region code makes sense to ensure drivers don't try to claim > (I/O) memory region claimed by another driver or by the memory > allocator. It doesn't make as much sense for registering available > memory banks, because there is often overlap: > > - device tree often defines the (minimally) available /memory across > variants > - subarchitectures like i.MX, STM32MP, OMAP or SAMA5 query the main > RAM size from controller registers > > So far, overlap went mostly unnoticed, because the RAM controller > drivers didn't check for errors. However, if barebox is already running > from RAM outside that described by the device tree, there will be errors > and /delete-node/ annotations that need to be added into device trees. > > Upstream device tree can be extended later on with a /memory node, which > in turn will break barebox. This series fixes all of that. Overlapping > memory banks are combined into one. Errors are propagated everywhere. > /delete-node/ memory@X shouldn't be needed for new boards. > > Ahmad Fatoum (6): > common: memory: allocate all memory devices at once > memory: fuse overlapping memory banks > of: propagate errors inside barebox_register_{of,fdt} into initcalls > of: warn about of_add_memory_bank errors > ARM: <asm/memory.h>: propagate error codes from arm_add_mem_device() > ARM: report probe error at arm_add_mem_device() callsites on failure Applied, thanks Sascha -- Pengutronix e.K. | | Steuerwalder Str. 21 | http://www.pengutronix.de/ | 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