On 08/06/2015 02:46 PM, Vlastimil Babka wrote:
On 07/17/2015 06:51 AM, Juergen Gross wrote:During early boot as Xen pv domain the kernel needs to map some page tables supplied by the hypervisor read only. This is needed to be able to relocate some data structures conflicting with the physical memory map especially on systems with huge RAM (above 512GB). Provide the function early_memremap_ro() to provide this read only mapping. Signed-off-by: Juergen Gross <jgross@xxxxxxxx> Acked-by: Konrad Rzeszutek Wilk <Konrad.wilk@xxxxxxxxxx> Cc: Arnd Bergmann <arnd@xxxxxxxx> Cc: linux-mm@xxxxxxxxx Cc: linux-arch@xxxxxxxxxxxxxxx --- include/asm-generic/early_ioremap.h | 2 ++ include/asm-generic/fixmap.h | 3 +++ mm/early_ioremap.c | 12 ++++++++++++ 3 files changed, 17 insertions(+) diff --git a/include/asm-generic/early_ioremap.h b/include/asm-generic/early_ioremap.h index a5de55c..316bd04 100644 --- a/include/asm-generic/early_ioremap.h +++ b/include/asm-generic/early_ioremap.h @@ -11,6 +11,8 @@ extern void __iomem *early_ioremap(resource_size_t phys_addr, unsigned long size); extern void *early_memremap(resource_size_t phys_addr, unsigned long size); +extern void *early_memremap_ro(resource_size_t phys_addr, + unsigned long size);So the function is declared unconditionally...extern void early_iounmap(void __iomem *addr, unsigned long size); extern void early_memunmap(void *addr, unsigned long size); diff --git a/include/asm-generic/fixmap.h b/include/asm-generic/fixmap.h index f23174f..1cbb833 100644 --- a/include/asm-generic/fixmap.h +++ b/include/asm-generic/fixmap.h @@ -46,6 +46,9 @@ static inline unsigned long virt_to_fix(const unsigned long vaddr) #ifndef FIXMAP_PAGE_NORMAL #define FIXMAP_PAGE_NORMAL PAGE_KERNEL #endif +#if !defined(FIXMAP_PAGE_RO) && defined(PAGE_KERNEL_RO) +#define FIXMAP_PAGE_RO PAGE_KERNEL_RO +#endif #ifndef FIXMAP_PAGE_NOCACHE #define FIXMAP_PAGE_NOCACHE PAGE_KERNEL_NOCACHE #endif diff --git a/mm/early_ioremap.c b/mm/early_ioremap.c index e10ccd2..0cfadaf 100644 --- a/mm/early_ioremap.c +++ b/mm/early_ioremap.c @@ -217,6 +217,13 @@ early_memremap(resource_size_t phys_addr, unsigned long size) return (__force void *)__early_ioremap(phys_addr, size, FIXMAP_PAGE_NORMAL); } +#ifdef FIXMAP_PAGE_RO +void __init * +early_memremap_ro(resource_size_t phys_addr, unsigned long size) +{ + return (__force void *)__early_ioremap(phys_addr, size, FIXMAP_PAGE_RO); +} +#endif... here we provide a implementation when both CONFIG_MMU and FIXMAP_PAGE_RO are defined...#else /* CONFIG_MMU */ void __init __iomem * @@ -231,6 +238,11 @@ early_memremap(resource_size_t phys_addr, unsigned long size) { return (void *)phys_addr; } +void __init * +early_memremap_ro(resource_size_t phys_addr, unsigned long size) +{ + return (void *)phys_addr; +}... and here for !CONFIG_MMU. So, what about CONFIG_MMU && !FIXMAP_PAGE_RO combinations? Which translates to CONFIG_MMU && !PAGE_KERNEL_RO. Maybe they don't exist, but then it's still awkward to see the combination in the code left unimplemented.
At least there are some architectures without #define PAGE_KERNEL_RO but testing CONFIG_MMU (arm, m68k, xtensa).
Would it be perhaps simpler to assume the same thing as in drivers/base/firmware_class.c ? /* Some architectures don't have PAGE_KERNEL_RO */ #ifndef PAGE_KERNEL_RO #define PAGE_KERNEL_RO PAGE_KERNEL #endif Or would it be dangerous here to silently lose the read-only protection?
The only reason to use this function instead of early_memremap() is the mandatory read-only mapping. My intention was to let the build fail in case it is being used but not implemented. An architecture requiring the function but having no PAGE_KERNEL_RO still can define FIXMAP_PAGE_RO. Juergen -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>