Tegra,Tegra124: proposed layout for file hierarchy with example
This change shows the source structure for nvidia Tegra and Tegra124
SOC. The problem we are trying to solve is that there is a large
amount of common code in the form of .c and .h files across many
different Tegra SOCs. The solution is to provide common code in a
single directory, but not to compile in the common code directory;
rather, we compile in a directory for a given SOC. Different SOCs
will sometimes need different bits of code from the common directory.
Tegra common code lives in tegra/, but there is no makefile there: if
a Tegra common file is needed in a SOC, it is referenced via a
Makefile in a specific Tegra SOC.
Another issue is includes. Include files in the common directory might be
accessed by a piece of code in an SOC directory. More problematically,
code in the common directory might require a file in an SOC directory.
We don't want to put the SOC name in an #include path, e.g.
in a C file in tegra/ is very undesirable, since we might be compiling
for a tegra114.
On some systems this is solved by a pre-pass which creates a set of
symbolic links; on others with nested #ifdef in the common code
which include different .h files depending on CPP variables.
In previous years, both LinuxBIOS and coreboot have tried these
solutions and found them inconvenient and error-prone.
We choose to solve it by requiring explicit naming of part of the path
of files that are in the common directory. This requirement, coupled
with two -I directives in the Makefile.inc, allows common and SOC
C code to incorporate both common and SOC .h files.
.c and .h files -- SOC or common -- name include
files in the common directory with the prefix tegra/, e.g.
SOC files will be included from the SOC directory if they have no prefix:
The full patch of clock.h will depend on what SOC is being compiled, which
is desirable.
In this way, a common file can pick up a specific SOC file without
creating symlinks or other such tricky magic.
We show this usage with one file, soc/nvidia/tega124/clock.c. This compiles.
The last question is where to put the prototype for the function
defined in this file -- soc.h?
Change-Id: Iecb635cec70f24a5b3e18caeda09d04a00d29409
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Reviewed-on: https://chromium-review.googlesource.com/171569
Reviewed-by: Ronald Minnich <rminnich@chromium.org>
Tested-by: Gabe Black <gabeblack@chromium.org>
Commit-Queue: Gabe Black <gabeblack@chromium.org>
(cherry picked from commit 53e3bed868953f3da588ec90661d316a6482e27e)
Signed-off-by: Isaac Christensen <isaac.christensen@se-eng.com>
Reviewed-on: http://review.coreboot.org/6621
Tested-by: build bot (Jenkins)
diff --git a/src/soc/nvidia/tegra124/Makefile.inc b/src/soc/nvidia/tegra124/Makefile.inc
index 14707df..0037a69 100644
--- a/src/soc/nvidia/tegra124/Makefile.inc
+++ b/src/soc/nvidia/tegra124/Makefile.inc
@@ -1,6 +1,9 @@
CBOOTIMAGE = cbootimage
bootblock-y += cbfs.c
+bootblock-y += clock.c
+bootblock-y += monotonic_timer.c
+bootblock-y += timer.c
romstage-y += cbfs.c
romstage-y += monotonic_timer.c
@@ -10,6 +13,8 @@
ramstage-y += monotonic_timer.c
ramstage-y += timer.c
+INCLUDES += -Isrc/soc/nvidia/tegra124 -Isrc/soc/nvidia
+
# We want to grab the bootblock right before it goes into the image and wrap
# it inside a BCT, but ideally we would do that without making special, one
# use modifications to the main ARM Makefile. We do this in two ways. First,