Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S932293AbdGCQvZ (ORCPT ); Mon, 3 Jul 2017 12:51:25 -0400 Received: from smtprelay.synopsys.com ([198.182.47.9]:53907 "EHLO smtprelay.synopsys.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754902AbdGCQvW (ORCPT ); Mon, 3 Jul 2017 12:51:22 -0400 From: Eugeniy Paltsev To: linux-snps-arc@lists.infradead.org Cc: linux-kernel@vger.kernel.org, Vineet Gupta , Alexey Brodkin , Rob Herring , devicetree@vger.kernel.org, Eugeniy Paltsev Subject: [PATCH 2/3 v7] ARC: Decouple linux kernel memory address and link address Date: Mon, 3 Jul 2017 19:51:08 +0300 Message-Id: <20170703165109.31535-3-Eugeniy.Paltsev@synopsys.com> X-Mailer: git-send-email 2.9.3 In-Reply-To: <20170703165109.31535-1-Eugeniy.Paltsev@synopsys.com> References: <20170703165109.31535-1-Eugeniy.Paltsev@synopsys.com> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 3985 Lines: 107 We faced with problem when we tried to utilize 1G DRAM by linux on HSDK. We can't use our usual kernel memory address (0x80000000) like on AXS103 because of DCCM memory bank located at exactly same address (0x80000000) But we can't simply move kernel memory address to another address (like 0x90000000) because IOC base address must be aligned to the size of the aperture as specified in the IOC size register. So we had to use 1G aligned address for kernel memory. We can't use 0x00000000 or 0x40000000 addresses because addresses lover then 0x80000000 are MMU-translated. We can't use 0xB0000000 address because we can define a volatile uncached region only from AUX_NON_VOLATILE_LIMIT to the 0xFFFFFFFF. (the end of region is hardcoded) So, the decision is to link kernel to 0x90000000, but use 0x80000000-0xBFFFFFFF memory region and reallocate DCCM in our platform code. This patch only makes possible to set kernel memory address not equal to kernel link address. Signed-off-by: Eugeniy Paltsev --- arch/arc/Kconfig | 5 +++++ arch/arc/include/asm/page.h | 2 +- arch/arc/mm/cache.c | 2 +- arch/arc/mm/init.c | 6 +++--- 4 files changed, 10 insertions(+), 5 deletions(-) diff --git a/arch/arc/Kconfig b/arch/arc/Kconfig index a545969..75e5276 100644 --- a/arch/arc/Kconfig +++ b/arch/arc/Kconfig @@ -430,6 +430,11 @@ config LINUX_LINK_BASE However some customers have peripherals mapped at this addr, so Linux needs to be scooted a bit. If you don't know what the above means, leave this setting alone. + +config KERNEL_RAM_BASE_ADDRESS + hex "Linux ram base address" + default LINUX_LINK_BASE + help This needs to match memory start address specified in Device Tree config HIGHMEM diff --git a/arch/arc/include/asm/page.h b/arch/arc/include/asm/page.h index 296c342..777f676 100644 --- a/arch/arc/include/asm/page.h +++ b/arch/arc/include/asm/page.h @@ -85,7 +85,7 @@ typedef pte_t * pgtable_t; */ #define virt_to_pfn(kaddr) (__pa(kaddr) >> PAGE_SHIFT) -#define ARCH_PFN_OFFSET virt_to_pfn(CONFIG_LINUX_LINK_BASE) +#define ARCH_PFN_OFFSET virt_to_pfn(CONFIG_KERNEL_RAM_BASE_ADDRESS) #ifdef CONFIG_FLATMEM #define pfn_valid(pfn) (((pfn) - ARCH_PFN_OFFSET) < max_mapnr) diff --git a/arch/arc/mm/cache.c b/arch/arc/mm/cache.c index 383ff77..f303274 100644 --- a/arch/arc/mm/cache.c +++ b/arch/arc/mm/cache.c @@ -1113,7 +1113,7 @@ noinline void __init arc_ioc_setup(void) * For now we assume IOC aperture to cover all the memory used by the * kernel. */ - ap_base = CONFIG_LINUX_LINK_BASE; + ap_base = CONFIG_KERNEL_RAM_BASE_ADDRESS; if (ap_base % ap_size != 0) panic("IOC Aperture start must be aligned to the size of the aperture"); diff --git a/arch/arc/mm/init.c b/arch/arc/mm/init.c index 8c9415e..f84cba2 100644 --- a/arch/arc/mm/init.c +++ b/arch/arc/mm/init.c @@ -26,7 +26,7 @@ pgd_t swapper_pg_dir[PTRS_PER_PGD] __aligned(PAGE_SIZE); char empty_zero_page[PAGE_SIZE] __aligned(PAGE_SIZE); EXPORT_SYMBOL(empty_zero_page); -static const unsigned long low_mem_start = CONFIG_LINUX_LINK_BASE; +static const unsigned long low_mem_start = CONFIG_KERNEL_RAM_BASE_ADDRESS; static unsigned long low_mem_sz; #ifdef CONFIG_HIGHMEM @@ -63,7 +63,7 @@ void __init early_init_dt_add_memory_arch(u64 base, u64 size) if (!low_mem_sz) { if (base != low_mem_start) - panic("CONFIG_LINUX_LINK_BASE != DT memory { }"); + panic("CONFIG_KERNEL_RAM_BASE_ADDRESS != DT memory { }"); low_mem_sz = size; in_use = 1; @@ -161,7 +161,7 @@ void __init setup_arch_memory(void) * We can't use the helper free_area_init(zones[]) because it uses * PAGE_OFFSET to compute the @min_low_pfn which would be wrong * when our kernel doesn't start at PAGE_OFFSET, i.e. - * PAGE_OFFSET != CONFIG_LINUX_LINK_BASE + * PAGE_OFFSET != CONFIG_KERNEL_RAM_BASE_ADDRESS */ free_area_init_node(0, /* node-id */ zones_size, /* num pages per zone */ -- 2.9.3