On Thu, 15 Oct 2015 16:42:19 +0200 Linus Walleij linus.walleij@linaro.org wrote:
On Fri, Oct 9, 2015 at 3:09 PM, Liviu Dudau Liviu.Dudau@foss.arm.com wrote:
Juno R1 has an XpressRICH3 PCIe host bridge that needs to be initialised in order for the Linux kernel to be able to enumerate the bus. Add support code here that enables the host bridge, trains the links and sets up the Address Translation Tables.
Signed-off-by: Liviu Dudau Liviu.Dudau@foss.arm.com
Very nice! Now we (soon) have PCIe on the Juno.
Did you:
Test with compiling in e.g. network cards and booting off of ethernet on PCIe?
Test what happens with a simple VGA card on PCIe? Sometimes the VGA card BIOS need to be initialized using an emulator running the x86 ROM and I never got that working on anything ARM :( (Maybe PCIe doesn't suffer from this? Just vanilla PCI has this problem? What do I know.)
Hi Linus,
Please note that this patchset does not enable enough functionality in U-Boot to allow for PCIe testing.
However, together with my Linux patchset series[1] I have tested U-Boot plus mainline Linux v4.3-rc5 on Juno r1 with 3x USB-to-PCIe cards and a Radeon HD6400 card. I have even got framebuffer out of the Radeon card :)
+#ifdef CONFIG_TARGET_VEXPRESS64_JUNO +void xr3pci_set_atr_entry(unsigned long base, unsigned long src_addr,
unsigned long trsl_addr, int window_size,int trsl_param)+{
(...)
int board_init(void) { +#ifdef CONFIG_TARGET_VEXPRESS64_JUNO
(...)
+#endif return 0; }
+++ b/board/armltd/vexpress64/vexpress64.h @@ -0,0 +1,60 @@ +#ifndef __VEXPRESS64_H__ +#define __VEXPRESS64_H__
Instead of peppering with #ifdefs I suggest you do like this:
Create two new files named board/armltd/vexpress64/pcie.c board/armltd/vexpress64/pcie.h
Move all the #defines from the vexpress64.h file into the top of the pcie.c file.
Use the pcie.h file for prototypes + stubs like this:
#ifdef CONFIG_TARGET_VEXPRESS64_JUNO void vexpress64_pcie_init(void); #else static inline void vexpress64_pcie_init(void) { } #endif
Then board_init() can unconditionallt call these functions and they will be stubbed out if compiled for anything else than Juno.
Yours, Linus Walleij
OK, I will do that. I'm waiting on some feedback from Ryan Harkin that wanted to test the patchset with some Linaro kernels and I will post a v2.
Best regards, Liviu