ioremap执行过程

驱动程序映射bar物理地址到虚拟地址时，通常使用ioremap系列函数。接下来，分析相关实现方式。

early_ioremap_init

  early_ioremap_init在start_kernel -> setup_arch函数中调用：

void __init early_ioremap_init(void) { pmd_t *pmd; #ifdef CONFIG_X86_64 BUILD_BUG_ON((fix_to_virt(0) + PAGE_SIZE) & ((1 << PMD_SHIFT) - 1)); // 检查低12位是否为0，为0继续执行 | | \/ /* #define __fix_to_virt(x) (FIXADDR_TOP - ((x) << PAGE_SHIFT)) , FIXADDR_TOP ：从 vsyscall空间的基址取整产生的地址，属于固定映射区域的最高地址*/ #else WARN_ON((fix_to_virt(0) + PAGE_SIZE) & ((1 << PMD_SHIFT) - 1)); #endif early_ioremap_setup(); // 初期固定映射地址填充slot_virt 数组 || \/ void __init early_ioremap_setup(void) { int i; for (i = 0; i < FIX_BTMAPS_SLOTS; i++) if (WARN_ON(prev_map[i])) // prev_map 数组用于记录初期 ioremap 区域的地址(临时启动时映射，实际上有512个，在ioremap可用之前使用)，这里还未使用，指向为0 /* static void __iomem *prev_map[FIX_BTMAPS_SLOTS] __initdata; * / /* #define __initdata __section(.init.data); __initdata数据存储在.init.data段，这些内存在内核初始化结束后释放掉 */ /* #define FIX_BTMAPS_SLOTS 8 */ break; for (i = 0; i < FIX_BTMAPS_SLOTS; i++) slot_virt[i] = __fix_to_virt(FIX_BTMAP_BEGIN - NR_FIX_BTMAPS*i); // 固定映射区域的虚拟地址 } 

  继续往下看：

pmd = early_ioremap_pmd(fix_to_virt(FIX_BTMAP_BEGIN)); || \/ static inline pmd_t * __init early_ioremap_pmd(unsigned long addr) { /* Don't assume we're using swapper_pg_dir at this point */ pgd_t *base = __va(read_cr3_pa()); 从cr3寄存器读出基地址 /* __read_cr3() & CR3_ADDR_MASK */ /* 屏蔽地址空间ID和SME加密位 */ pgd_t *pgd = &base[pgd_index(addr)]; // pgd页目录(page global directory) /* pgd_index(addr) -> #define pgd_index(a) (((a) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1)) */ /* #define PGDIR_SHIFT pgdir_shift */ /* unsigned int pgdir_shift __ro_after_init = 39; */ /* #define PTRS_PER_PGD 512 */ p4d_t *p4d = p4d_offset(pgd, addr); // 4级页目录 || \/ static inline p4d_t *p4d_offset(pgd_t *pgd, unsigned long address) { if (!pgtable_l5_enabled()) // 如果.config中没有开启CONFIG_X86_5LEVEL return (p4d_t *)pgd; return (p4d_t *)pgd_page_vaddr(*pgd) + p4d_index(address); // pgd + 4级页表偏移值(偏移值为0，目前的计算机内存4级表应该足够使用了) /* (pgd.pgd & PGD_ALLOWED_BITS) + (address >> P4D_SHIFT) & (PTRS_PER_P4D - 1) */ } /* #define PGD_ALLOWED_BITS (PGD_PAE_PHYS_MASK | _PAGE_PRESENT | \ _PAGE_PWT | _PAGE_PCD | \ _PAGE_SOFTW1 | _PAGE_SOFTW2 | _PAGE_SOFTW3) */ /* #define PTRS_PER_P4D ptrs_per_p4d */ /* unsigned int ptrs_per_p4d __ro_after_init = 1; */ /* define P4D_SHIFT 39 */ pud_t *pud = pud_offset(p4d, addr); // pud页目录(page upper directory) || \/ static inline pud_t *pud_offset(p4d_t *p4d, unsigned long address) { return p4d_pgtable(*p4d) + pud_index(address); // 偏移范围小于等于511 /* (pud_t *)__va(p4d_val(p4d) & p4d_pfn_mask(p4d)) + pud_index(address) */ } /* #define pud_index(x) (((x) >> PUD_SHIFT) & (PTRS_PER_PUD-1)) */ /* #define PUD_SHIFT 30 */ /* #define PTRS_PER_PUD 512 */ pmd_t *pmd = pmd_offset(pud, addr); // pmd页目录(page middle directory) || \/ static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address) { return pud_pgtable(*pud) + pmd_index(address); /* (pmd_t *)__va(pud_val(pud) & pud_pfn_mask(pud)) + pmd_index(address); } || \/ static inline pudval_t pud_pfn_mask(pud_t pud) { if (native_pud_val(pud) & _PAGE_PSE) // 4 MB or 2MB page return PHYSICAL_PUD_PAGE_MASK; // 值为 0000 0000 0000 0000 0000 0000 0000 0011 1111 1111 1111 1111 111 (31位) — (51位) else return PTE_PFN_MASK; // 值为 0000 0000 0000 1111 1111 1111 1111 1111 1111 1111 1111 1111 111 (12位) — (51位) } /* #define PHYSICAL_PUD_PAGE_MASK (((signed long)PUD_PAGE_MASK) & __PHYSICAL_MASK) */ /* #define PUD_PAGE_MASK (~(PUD_PAGE_SIZE-1)) */ /* #define PUD_PAGE_SIZE (_AC(1, UL) << PUD_SHIFT) */ /* #define __PHYSICAL_MASK physical_mask */ /* phys_addr_t physical_mask = (1ULL << __PHYSICAL_MASK_SHIFT) - 1; */ /* #else #define __PHYSICAL_MASK ((phys_addr_t)((1ULL << __PHYSICAL_MASK_SHIFT) - 1)) #endif */ /* #define __PHYSICAL_MASK_SHIFT 52 */ /* Extracts the PFN from a (pte|pmd|pud|pgd)val_t of a 4KB page */ /* #define PTE_PFN_MASK ((pteval_t)PHYSICAL_PAGE_MASK) */ /* #define PHYSICAL_PAGE_MASK (((signed long)PAGE_MASK) & __PHYSICAL_MASK) */ /* #define PAGE_MASK (~(PAGE_SIZE-1)) */ return pmd; } /* #ifdef CONFIG_DYNAMIC_MEMORY_LAYOUT unsigned long page_offset_base __ro_after_init = __PAGE_OFFSET_BASE_L4; EXPORT_SYMBOL(page_offset_base); unsigned long vmalloc_base __ro_after_init = __VMALLOC_BASE_L4; EXPORT_SYMBOL(vmalloc_base); unsigned long vmemmap_base __ro_after_init = __VMEMMAP_BASE_L4; EXPORT_SYMBOL(vmemmap_base); #endif #define __PAGE_OFFSET_BASE_L4 _AC(0xffff0, UL) #define __VMALLOC_BASE_L4 0xffffcUL #define __VMEMMAP_BASE_L4 0xffffea0000000000UL */ 

  继续往下看：

memset(bm_pte, 0, sizeof(bm_pte)); // bm_pte占一页大小 /* static pte_t bm_pte[PAGE_SIZE/sizeof(pte_t)] __page_aligned_bss; */ pmd_populate_kernel(&init_mm, pmd, bm_pte); // 为init_mm对象分配pte，并记录pmd || \/ static inline void pmd_populate_kernel(struct mm_struct *mm, pmd_t *pmd, pte_t *pte) { paravirt_alloc_pte(mm, __pa(pte) >> PAGE_SHIFT); // mmu分配pte -> mm /* PVOP_VCALL2(mmu.alloc_pte, mm, pfn) ; mmu.alloc_pte = pfn */ set_pmd(pmd, __pmd(__pa(pte) | _PAGE_TABLE)); // mmu设置pmd <- pmd /* PVOP_VCALL2(mmu.set_pmd, pmdp, native_pmd_val(pmd)); mmu.set_pmd = native_pmd_val(pmd) */ } /* #define __pa(x) __phys_addr((unsigned long)(x)) */ /* x = y + ((x > y) ? phys_base : (__START_KERNEL_map - PAGE_OFFSET)); */ /* #define __START_KERNEL_map _AC(0xffffffff, UL) */ if (pmd != early_ioremap_pmd(fix_to_virt(FIX_BTMAP_END))) { // 如果pmd不属于固定映射区域 WARN_ON(1); printk(KERN_WARNING "pmd %p != %p\n", pmd, early_ioremap_pmd(fix_to_virt(FIX_BTMAP_END))); printk(KERN_WARNING "fix_to_virt(FIX_BTMAP_BEGIN): %08lx\n", fix_to_virt(FIX_BTMAP_BEGIN)); printk(KERN_WARNING "fix_to_virt(FIX_BTMAP_END): %08lx\n", fix_to_virt(FIX_BTMAP_END)); printk(KERN_WARNING "FIX_BTMAP_END: %d\n", FIX_BTMAP_END); printk(KERN_WARNING "FIX_BTMAP_BEGIN: %d\n", FIX_BTMAP_BEGIN); } 

ioremap

void __iomem *ioremap(resource_size_t phys_addr, unsigned long size) { enum page_cache_mode pcm = _PAGE_CACHE_MODE_UC_MINUS; return __ioremap_caller(phys_addr, size, pcm, __builtin_return_address(0), false); } || \/ static void __iomem * __ioremap_caller(resource_size_t phys_addr, unsigned long size, enum page_cache_mode pcm, void *caller, bool encrypted) { ... __ioremap_check_mem(phys_addr, size, &io_desc); // 检查物理内存是否被标记，不涵盖EFI runtime services数据区域，但当SEV处于活动状态时，必须对其进行加密映射 if (io_desc.flags & IORES_MAP_SYSTEM_RAM) { // 不能映射RAM内存 WARN_ONCE(1, "ioremap on RAM at %pa - %pa\n", &phys_addr, &last_addr); return NULL; } offset = phys_addr & ~PAGE_MASK; phys_addr &= PHYSICAL_PAGE_MASK; size = PAGE_ALIGN(last_addr+1) - phys_addr; retval = memtype_reserve(phys_addr, (u64)phys_addr + size, pcm, &new_pcm); // 在rbtrees(红黑树)中嵌入节点 prot = PAGE_KERNEL_IO; if ((io_desc.flags & IORES_MAP_ENCRYPTED) || encrypted) // 如果正在映射的页面在内存中并且SEV处于活动状态，则设置内存加密。在TDX客户机中，默认情况下内存被标记为私有，如果加密未请求(已使用加密)，显式设置解密 *确保在中启用了内存加密属性 prot = pgprot_encrypted(prot); else prot = pgprot_decrypted(prot); area = get_vm_area_caller(size, VM_IOREMAP, caller); //执行vmap分配函数，设置虚拟内存区域 if (!area) goto err_free_memtype; area->phys_addr = phys_addr; vaddr = (unsigned long) area->addr; if (memtype_kernel_map_sync(phys_addr, size, pcm)) goto err_free_area; if (ioremap_page_range(vaddr, vaddr + size, phys_addr, prot)) // 将物理地址映射到虚拟地址 goto err_free_area; ret_addr = (void __iomem *) (vaddr + offset); mmiotrace_ioremap(unaligned_phys_addr, unaligned_size, ret_addr); // 注册trace if (iomem_map_sanity_check(unaligned_phys_addr, unaligned_size)) // 检查请求是否跨越iomem资源树中的任何条 pr_warn("caller %pS mapping multiple BARs\n", caller); return ret_addr; err_free_area: free_vm_area(area); err_free_memtype: memtype_free(phys_addr, phys_addr + size); return NULL; }