Message ID | 1456329708-2042-1-git-send-email-miles.chen@mediatek.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
On Thu, Feb 25, 2016 at 12:01:48AM +0800, miles.chen@mediatek.com wrote: > From: Miles Chen <miles.chen@mediatek.com> > > The MODULES_VADDR is not the lowest possible > kernel virtual address. TASK_SIZE_64 may be larger than > MODULES_VADDR, FIXADDR_TOP, and PCI_IO_START. > > Fix this by comparing TASK_SIZE_64 (highest user virtual address) > with VA_START (lowest kernel virtual address). > > #define VA_BITS (CONFIG_ARM64_VA_BITS) > #define VA_START (UL(0xffffffffffffffff) << VA_BITS) > #define PAGE_OFFSET (UL(0xffffffffffffffff) << (VA_BITS - 1)) > #define MODULES_END (PAGE_OFFSET) > #define MODULES_VADDR (MODULES_END - SZ_64M) > #define PCI_IO_END (MODULES_VADDR - SZ_2M) > #define PCI_IO_START (PCI_IO_END - PCI_IO_SIZE) > #define FIXADDR_TOP (PCI_IO_START - SZ_2M) > #define TASK_SIZE_64 (UL(1) << VA_BITS) We should remove these checks altogether. There is a huge gap between the user and kernel addresses that they would never overlap (we don't have enough levels of page tables to cover 64-bit VA space).
On Wed, 2016-02-24 at 16:19 +0000, Catalin Marinas wrote: > On Thu, Feb 25, 2016 at 12:01:48AM +0800, miles.chen@mediatek.com wrote: > > From: Miles Chen <miles.chen@mediatek.com> > > > > The MODULES_VADDR is not the lowest possible > > kernel virtual address. TASK_SIZE_64 may be larger than > > MODULES_VADDR, FIXADDR_TOP, and PCI_IO_START. > > > > Fix this by comparing TASK_SIZE_64 (highest user virtual address) > > with VA_START (lowest kernel virtual address). > > > > #define VA_BITS (CONFIG_ARM64_VA_BITS) > > #define VA_START (UL(0xffffffffffffffff) << VA_BITS) > > #define PAGE_OFFSET (UL(0xffffffffffffffff) << (VA_BITS - 1)) > > #define MODULES_END (PAGE_OFFSET) > > #define MODULES_VADDR (MODULES_END - SZ_64M) > > #define PCI_IO_END (MODULES_VADDR - SZ_2M) > > #define PCI_IO_START (PCI_IO_END - PCI_IO_SIZE) > > #define FIXADDR_TOP (PCI_IO_START - SZ_2M) > > #define TASK_SIZE_64 (UL(1) << VA_BITS) > > We should remove these checks altogether. There is a huge gap between > the user and kernel addresses that they would never overlap (we don't > have enough levels of page tables to cover 64-bit VA space). > Thanks for your reply. I confirmed that in the ARMv8 documentation. I'll send another patch to remove these checks. Cheers, Miles
diff --git a/arch/arm64/mm/init.c b/arch/arm64/mm/init.c index f3b061e..7d75697 100644 --- a/arch/arm64/mm/init.c +++ b/arch/arm64/mm/init.c @@ -343,8 +343,8 @@ void __init mem_init(void) #ifdef CONFIG_COMPAT BUILD_BUG_ON(TASK_SIZE_32 > TASK_SIZE_64); #endif - BUILD_BUG_ON(TASK_SIZE_64 > MODULES_VADDR); - BUG_ON(TASK_SIZE_64 > MODULES_VADDR); + BUILD_BUG_ON(TASK_SIZE_64 > VA_START); + BUG_ON(TASK_SIZE_64 > VA_START); if (PAGE_SIZE >= 16384 && get_num_physpages() <= 128) { extern int sysctl_overcommit_memory;