On Tue, Oct 20, 2015 at 08:05:40AM +0200, Linus Walleij wrote:
Only compile in PCIe support if the board really uses it. Provide a stub for the init function if e.g. FVP is being built.
Cc: Liviu Dudau Liviu.Dudau@foss.arm.com Cc: Ryan Harkin ryan.harkin@linaro.org Signed-off-by: Linus Walleij linus.walleij@linaro.org
board/armltd/vexpress64/Makefile | 3 ++- board/armltd/vexpress64/pcie.c | 2 -- board/armltd/vexpress64/pcie.h | 4 ++++ 3 files changed, 6 insertions(+), 3 deletions(-)
diff --git a/board/armltd/vexpress64/Makefile b/board/armltd/vexpress64/Makefile index a35db401b684..b4391a71249a 100644 --- a/board/armltd/vexpress64/Makefile +++ b/board/armltd/vexpress64/Makefile @@ -5,4 +5,5 @@ # SPDX-License-Identifier: GPL-2.0+ #
-obj-y := vexpress64.o pcie.o +obj-y := vexpress64.o +obj-$(CONFIG_TARGET_VEXPRESS64_JUNO) += pcie.o diff --git a/board/armltd/vexpress64/pcie.c b/board/armltd/vexpress64/pcie.c index 7b999e8ef40b..311c4500e3ff 100644 --- a/board/armltd/vexpress64/pcie.c +++ b/board/armltd/vexpress64/pcie.c @@ -191,7 +191,5 @@ void xr3pci_init(void)
void vexpress64_pcie_init(void) { -#ifdef CONFIG_TARGET_VEXPRESS64_JUNO xr3pci_init(); -#endif } diff --git a/board/armltd/vexpress64/pcie.h b/board/armltd/vexpress64/pcie.h index 14642f4f5c43..55b276d6af11 100644 --- a/board/armltd/vexpress64/pcie.h +++ b/board/armltd/vexpress64/pcie.h @@ -1,6 +1,10 @@ #ifndef __VEXPRESS64_PCIE_H__ #define __VEXPRESS64_PCIE_H__
+#ifdef CONFIG_TARGET_VEXPRESS64_JUNO void vexpress64_pcie_init(void); +#else +static inline void vexpress64_pcie_init(void) {} +#endif
#endif /* __VEXPRESS64_PCIE_H__ */
Alright, so the versatile platform makes life fun at times. First, my preference is for weak functions (and I _think_ the linker ends up being smart enough about them to optimize things away, if not, arrg). Second, the question I have is, is this xr3pci_init() bit really a Juno thing or a baseboard thing (I assume no) or a going to be the same on the next Cortex-Asomething module thing? The way the code stands today a hypothethical A72-based vexpress64 module would need to call vexpress64_pcie_init() but not strictly xr3pci_init(). It would be a little ugly as pcie.c would #if/#elif/#endif the enabler call however so we could __weak a no-op in board.c and rename pcie.c to juno.c and have a72codenameboard.c later on with its own function.