Hi Wolfram, On 29 October 2015 12:38, Wolfram wrote: > > Fixes and changes to get PCIe working on ARM64 with mulitple instances. > > > I've tested these on ARM (Koelsch board), and it works fine. > > I've also tested on ARM64 (Salvator-X board), but I currently have an issue > > I wanted to test these patches on arm64 as well, but even after merging > pci-next, I still get: > > drivers/pci/host/pcie-rcar.c: In function 'rcar_pcie_setup': > drivers/pci/host/pcie-rcar.c:380:4: error: implicit declaration of function > 'pci_ioremap_io' [-Werror=implicit-function-declaration] > > So, what do I miss? Ouch, my bad. I have been working with our out-of-tree BSP for the Salvator-X board on http://git.kernel.org/cgit/linux/kernel/git/horms/renesas-bsp.git, rcar-3.0.2 tag along with some patches from colleagues. I hadn't noticed this patch: >From 913ff73dcf19bb57b0a7ad69e51105560d69fa9b Mon Sep 17 00:00:00 2001 From: Harunobu Kurokawa <harunobu.kurokawa.dn@xxxxxxxxxxx> Date: Thu, 24 Sep 2015 22:55:58 +0900 Subject: [PATCH 03/15] arm64: PCI(e) arch support for ioremap function Refered the followings patch: http://lists.infradead.org/pipermail/linux-arm-kernel/2014-July/269206.html Signed-off-by: Harunobu Kurokawa <harunobu.kurokawa.dn@xxxxxxxxxxx> --- arch/arm64/include/asm/io.h | 2 ++ arch/arm64/mm/ioremap.c | 13 +++++++++++++ 2 files changed, 15 insertions(+) diff --git a/arch/arm64/include/asm/io.h b/arch/arm64/include/asm/io.h index 44be1e0..cea3651 100644 --- a/arch/arm64/include/asm/io.h +++ b/arch/arm64/include/asm/io.h @@ -149,6 +149,8 @@ static inline u64 __raw_readq(const volatile void __iomem *addr) #define IO_SPACE_LIMIT (PCI_IO_SIZE - 1) #define PCI_IOBASE ((void __iomem *)PCI_IO_START) +extern int pci_ioremap_io(unsigned int offset, phys_addr_t phys_addr); + /* * String version of I/O memory access operations. */ diff --git a/arch/arm64/mm/ioremap.c b/arch/arm64/mm/ioremap.c index 01e88c8..34f09a6 100644 --- a/arch/arm64/mm/ioremap.c +++ b/arch/arm64/mm/ioremap.c @@ -104,6 +104,19 @@ void __iomem *ioremap_cache(phys_addr_t phys_addr, size_t size) } EXPORT_SYMBOL(ioremap_cache); +#ifdef CONFIG_PCI +int pci_ioremap_io(unsigned int offset, phys_addr_t phys_addr) +{ + BUG_ON(offset + SZ_64K > IO_SPACE_LIMIT); + + return ioremap_page_range((unsigned long)PCI_IOBASE + offset, + (unsigned long)PCI_IOBASE + offset + SZ_64K, + phys_addr, + __pgprot(PROT_NORMAL_NC)); +} +EXPORT_SYMBOL_GPL(pci_ioremap_io); +#endif + /* * Must be called after early_fixmap_init */ -- 1.7.9.5 Additionally, I have found out that MSI on R-Car PCIe upstream is broken. You can workaround this by reverting f075915ac0b11847fcfc8c4d55526a317e71c4d1 and f075915ac0b11847fcfc8c4d55526a317e71c4d1 until we fix this. > > with inbound PCI accesses. I am reasonably sure that this problem is hardware > > related. > > Any news if this is really HW related? I found out that the PCIe controller HW only supports 32-bit AXI bus addresses, so it works fine if you limit the available memory. The driver will need some work to use the IPMMU at some point. Thanks Phil -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
