This is an automated email from the ASF dual-hosted git repository. xiaoxiang pushed a commit to branch master in repository https://gitbox.apache.org/repos/asf/nuttx.git
commit 0200a48f7260bbaf596c77f457a50efefb2a0b18 Author: p-szafonimateusz <[email protected]> AuthorDate: Mon Jun 17 14:05:13 2024 +0200 arch/intel64/intel64_map_region.c: use MMU api to implement up_map_region use new MMU api to implement up_map_region(). The new implementation support maping over 0xffffffff but requires CONFIG_MM_PGALLOC=y Signed-off-by: p-szafonimateusz <[email protected]> --- arch/x86_64/include/intel64/arch.h | 3 +- arch/x86_64/src/intel64/intel64_map_region.c | 80 +++++++++++++++++++++++++++- 2 files changed, 80 insertions(+), 3 deletions(-) diff --git a/arch/x86_64/include/intel64/arch.h b/arch/x86_64/include/intel64/arch.h index 7539efaff8..8be9273e3e 100644 --- a/arch/x86_64/include/intel64/arch.h +++ b/arch/x86_64/include/intel64/arch.h @@ -34,6 +34,7 @@ #ifndef __ASSEMBLY__ # include <nuttx/compiler.h> # include <stdint.h> +# include <stddef.h> #endif /**************************************************************************** @@ -494,7 +495,7 @@ extern volatile struct gdt_entry_s *g_gdt64; * Public Function Prototypes ****************************************************************************/ -int up_map_region(void *base, int size, int flags); +int up_map_region(void *base, size_t size, int flags); void x86_64_check_and_enable_capability(void); extern void __enable_sse_avx(void); diff --git a/arch/x86_64/src/intel64/intel64_map_region.c b/arch/x86_64/src/intel64/intel64_map_region.c index 69fd2ce8b8..40f35622d6 100644 --- a/arch/x86_64/src/intel64/intel64_map_region.c +++ b/arch/x86_64/src/intel64/intel64_map_region.c @@ -26,8 +26,11 @@ #include <debug.h> #include <nuttx/irq.h> +#include <nuttx/pgalloc.h> #include "x86_64_internal.h" +#include "x86_64_mmu.h" +#include "pgalloc.h" /**************************************************************************** * Public Functions @@ -41,7 +44,77 @@ * ****************************************************************************/ -int up_map_region(void *base, int size, int flags) +#ifdef CONFIG_MM_PGALLOC +int up_map_region(void *base, size_t size, int flags) +{ + uintptr_t bb; + int ptlevel; + uintptr_t ptprev; + uintptr_t paddr; + uintptr_t vaddr; + size_t nmapped; + int i; + + /* Round to page boundary */ + + bb = (uintptr_t)base & ~(PAGE_SIZE - 1); + + /* Increase size if the base address is rounded off */ + + size += (uintptr_t)base - bb; + + /* Map 1:1 */ + + vaddr = bb; + nmapped = 0; + + while (nmapped < size) + { + /* Start from PTL4 */ + + ptprev = x86_64_pgvaddr(get_pml4()); + + for (ptlevel = 0; ptlevel < X86_MMU_PT_LEVELS - 1; ptlevel++) + { + paddr = mmu_pte_to_paddr(mmu_ln_getentry(ptlevel, ptprev, vaddr)); + if (!paddr) + { + /* Nothing yet, allocate one page for final level page table */ + + paddr = mm_pgalloc(1); + if (!paddr) + { + return -ENOMEM; + } + + /* Map the page table to the prior level */ + + mmu_ln_setentry(ptlevel, ptprev, paddr, vaddr, 0); + + /* This is then used to map the final level */ + + x86_64_pgwipe(paddr); + } + + ptprev = x86_64_pgvaddr(paddr); + } + + /* Then map the virtual address to the physical address */ + + for (i = X86_MMU_VADDR_INDEX(vaddr, ptlevel); + i < X86_MMU_ENTRIES_PER_PGT && nmapped < size; + i++) + { + mmu_ln_setentry(ptlevel, ptprev, bb + nmapped, vaddr, flags); + nmapped += MM_PGSIZE; + vaddr += MM_PGSIZE; + } + } + + return 0; +} +#else +int up_map_region(void *base, size_t size, int flags) { uint64_t bb; uint64_t num_of_pages; @@ -60,7 +133,9 @@ int up_map_region(void *base, int size, int flags) if (bb > 0xffffffff) { - return -1; /* Only < 4GB can be mapped */ + /* More than 4GB can't be mapped with this implementtion */ + + PANIC(); } curr = bb; @@ -74,3 +149,4 @@ int up_map_region(void *base, int size, int flags) return 0; } +#endif
