Cc: Bjorn Helgaas <bjorn.helgaas@...>, Jesse Barnes <jbarnes@...>, Len Brown <lenb@...>, Frans Pop <elendil@...>, Rafael J. Wysocki <rjw@...>, Linux Kernel Mailing List <linux-kernel@...>, <linux-pci@...>, <linux-acpi@...>, Adam Belay <abelay@...>, Avuton Olrich <avuton@...>, Karl Bellve <karl.bellve@...>, Willem Riede <wriede@...>, Matthew Hall <mhall@...>, Sam Ravnborg <sam@...>
A clearer (and more flexible) solution might be this patch. It's more
explicit about the fact that it simply makes it clear that any drivers
that are added by the architecture Makefile will be _first_ in the list of
drivers.
I suspect we should do the same with 'libs' and 'core' too, for that
matter, but we don't really use '*.a' libraries any more so link order
doesn't matter apart from initcall ordering. And libraries should hopfully
never have that issue. And 'core' is at least right now just the
initramfs thing. So in practice, it's probably only drivers/ that coul
have issues like this.
Sam added to Cc, since this is a build issue. What do people think?
Background: We are very careful to add 'drivers/pci' _before_
'drivers/{acpi,pnp}' in the drivers/Makefile, but what happens now is that
since the arch Makefile adds it's own drivers to the very end, the really
core PCI initcalls actually get ordered at the end, not the beginning.
# Objects we will link into vmlinux / subdirs we need to visit
init-y := init/
-drivers-y := drivers/ sound/ firmware/
net-y := net/
libs-y := lib/
core-y := usr/
@@ -517,6 +516,9 @@ endif
include $(srctree)/arch/$(SRCARCH)/Makefile
+# Do this _after_ the architecture may have added its own core drivers
+drivers-y += drivers/ sound/ firmware/
+
ifneq (CONFIG_FRAME_WARN,0)
KBUILD_CFLAGS += $(call cc-option,-Wframe-larger-than=${CONFIG_FRAME_WARN})
endif
--
