Hi TF, first, thanks for taking time to continue having the discussion around this patch series.
On Tue, May 21, 2019 at 05:40:59AM +0000, Chee, Tien Fong wrote: > On Thu, 2019-05-16 at 15:54 -0500, Andreas Dannenberg wrote: > > To give us more flexibility using the FS loader eliminate the need of > > always having to use the ENV to configure the block device but rather > > allow the respective block device and partition to be setup through > > platform data. > > > > Signed-off-by: Andreas Dannenberg <dannenb...@ti.com> > > Why not using DT method? We would like to use a single U-Boot build to support all of the different boot modes (boot media) a given device supports. In case of this RFC series this is really distilled down to only mmc0 and mmc1 but this already illustrates the point. The patch 05/11 shows how spl_boot_device() is used to make the MMC probing dependent on the boot device mmc0 vs mmc1. So hard-coding in DT will lose this flexibility. Of course there would be ways out such as dynamically updating DTS or to extend the FS loader probe function to select from multiple DTS nodes based on boot device, but all would come at a cost of additional memory consumption. -- Andreas Dannenberg Texas Instruments Inc _______________________________________________ U-Boot mailing list U-Boot@lists.denx.de https://lists.denx.de/listinfo/u-boot
