Hi Ben, > > When we're using CONFIG_ARM_DEVTREE, we need to postpone machine > > detection until later in setup_arch. Because ARM_DEVTREE depends on > > !DEBUG_LL, we don't need the mdesc this early anyway. > > > > We'll add support for ARM_DEVTREE && DEBUG_LL later. > > Why not add a specific machine type for 'this is device tree' and > then these routines don't even need to #ifdef. It would especially > be good as it allows dt and non-dt machines to coexist in the same > kernel together.
Mainly because this doesn't gain us much; the machine_desc that is discovered from this new machine type would not have the correct data (phys_io & io_pg_offset) required for early setup (used for DEBUG_LL). We could conditionally skip the detection if we see a DT machine type, but then we'd need to add similar checks for all the paths where the machine_desc is used. However, my goal is indeed to allow DT and non-DT machines to coexist as much as possible, this patch is more a temporary workaround while I figure out how to handle the DEBUG_LL stuff nicely for DT cases. Cheers, Jeremy _______________________________________________ devicetree-discuss mailing list [email protected] https://lists.ozlabs.org/listinfo/devicetree-discuss
