1 /*
2 * prepare to run common code
3 *
4 * Copyright (C) 2000 Andrea Arcangeli <andrea@suse.de> SuSE
5 */
6
7 #include <linux/init.h>
8 #include <linux/linkage.h>
9 #include <linux/types.h>
10 #include <linux/kernel.h>
11 #include <linux/string.h>
12 #include <linux/percpu.h>
13 #include <linux/start_kernel.h>
14 #include <linux/io.h>
15 #include <linux/memblock.h>
16
17 #include <asm/processor.h>
18 #include <asm/proto.h>
19 #include <asm/smp.h>
20 #include <asm/setup.h>
21 #include <asm/desc.h>
22 #include <asm/pgtable.h>
23 #include <asm/tlbflush.h>
24 #include <asm/sections.h>
25 #include <asm/kdebug.h>
26 #include <asm/e820.h>
27 #include <asm/bios_ebda.h>
28 #include <asm/bootparam_utils.h>
29 #include <asm/microcode.h>
30
31 /*
32 * Manage page tables very early on.
33 */
34 extern pgd_t early_level4_pgt[PTRS_PER_PGD];
35 extern pmd_t early_dynamic_pgts[EARLY_DYNAMIC_PAGE_TABLES][PTRS_PER_PMD];
36 static unsigned int __initdata next_early_pgt = 2;
37 pmdval_t early_pmd_flags = __PAGE_KERNEL_LARGE & ~(_PAGE_GLOBAL | _PAGE_NX);
38
39 /* Wipe all early page tables except for the kernel symbol map */
reset_early_page_tables(void)40 static void __init reset_early_page_tables(void)
41 {
42 unsigned long i;
43
44 for (i = 0; i < PTRS_PER_PGD-1; i++)
45 early_level4_pgt[i].pgd = 0;
46
47 next_early_pgt = 0;
48
49 write_cr3(__pa(early_level4_pgt));
50 }
51
52 /* Create a new PMD entry */
early_make_pgtable(unsigned long address)53 int __init early_make_pgtable(unsigned long address)
54 {
55 unsigned long physaddr = address - __PAGE_OFFSET;
56 unsigned long i;
57 pgdval_t pgd, *pgd_p;
58 pudval_t pud, *pud_p;
59 pmdval_t pmd, *pmd_p;
60
61 /* Invalid address or early pgt is done ? */
62 if (physaddr >= MAXMEM || read_cr3() != __pa(early_level4_pgt))
63 return -1;
64
65 again:
66 pgd_p = &early_level4_pgt[pgd_index(address)].pgd;
67 pgd = *pgd_p;
68
69 /*
70 * The use of __START_KERNEL_map rather than __PAGE_OFFSET here is
71 * critical -- __PAGE_OFFSET would point us back into the dynamic
72 * range and we might end up looping forever...
73 */
74 if (pgd)
75 pud_p = (pudval_t *)((pgd & PTE_PFN_MASK) + __START_KERNEL_map - phys_base);
76 else {
77 if (next_early_pgt >= EARLY_DYNAMIC_PAGE_TABLES) {
78 reset_early_page_tables();
79 goto again;
80 }
81
82 pud_p = (pudval_t *)early_dynamic_pgts[next_early_pgt++];
83 for (i = 0; i < PTRS_PER_PUD; i++)
84 pud_p[i] = 0;
85 *pgd_p = (pgdval_t)pud_p - __START_KERNEL_map + phys_base + _KERNPG_TABLE;
86 }
87 pud_p += pud_index(address);
88 pud = *pud_p;
89
90 if (pud)
91 pmd_p = (pmdval_t *)((pud & PTE_PFN_MASK) + __START_KERNEL_map - phys_base);
92 else {
93 if (next_early_pgt >= EARLY_DYNAMIC_PAGE_TABLES) {
94 reset_early_page_tables();
95 goto again;
96 }
97
98 pmd_p = (pmdval_t *)early_dynamic_pgts[next_early_pgt++];
99 for (i = 0; i < PTRS_PER_PMD; i++)
100 pmd_p[i] = 0;
101 *pud_p = (pudval_t)pmd_p - __START_KERNEL_map + phys_base + _KERNPG_TABLE;
102 }
103 pmd = (physaddr & PMD_MASK) + early_pmd_flags;
104 pmd_p[pmd_index(address)] = pmd;
105
106 return 0;
107 }
108
109 /* Don't add a printk in there. printk relies on the PDA which is not initialized
110 yet. */
clear_bss(void)111 static void __init clear_bss(void)
112 {
113 memset(__bss_start, 0,
114 (unsigned long) __bss_stop - (unsigned long) __bss_start);
115 }
116
get_cmd_line_ptr(void)117 static unsigned long get_cmd_line_ptr(void)
118 {
119 unsigned long cmd_line_ptr = boot_params.hdr.cmd_line_ptr;
120
121 cmd_line_ptr |= (u64)boot_params.ext_cmd_line_ptr << 32;
122
123 return cmd_line_ptr;
124 }
125
copy_bootdata(char * real_mode_data)126 static void __init copy_bootdata(char *real_mode_data)
127 {
128 char * command_line;
129 unsigned long cmd_line_ptr;
130
131 memcpy(&boot_params, real_mode_data, sizeof boot_params);
132 sanitize_boot_params(&boot_params);
133 cmd_line_ptr = get_cmd_line_ptr();
134 if (cmd_line_ptr) {
135 command_line = __va(cmd_line_ptr);
136 memcpy(boot_command_line, command_line, COMMAND_LINE_SIZE);
137 }
138 }
139
x86_64_start_kernel(char * real_mode_data)140 asmlinkage __visible void __init x86_64_start_kernel(char * real_mode_data)
141 {
142 int i;
143
144 /*
145 * Build-time sanity checks on the kernel image and module
146 * area mappings. (these are purely build-time and produce no code)
147 */
148 BUILD_BUG_ON(MODULES_VADDR < __START_KERNEL_map);
149 BUILD_BUG_ON(MODULES_VADDR - __START_KERNEL_map < KERNEL_IMAGE_SIZE);
150 BUILD_BUG_ON(MODULES_LEN + KERNEL_IMAGE_SIZE > 2*PUD_SIZE);
151 BUILD_BUG_ON((__START_KERNEL_map & ~PMD_MASK) != 0);
152 BUILD_BUG_ON((MODULES_VADDR & ~PMD_MASK) != 0);
153 BUILD_BUG_ON(!(MODULES_VADDR > __START_KERNEL));
154 BUILD_BUG_ON(!(((MODULES_END - 1) & PGDIR_MASK) ==
155 (__START_KERNEL & PGDIR_MASK)));
156 BUILD_BUG_ON(__fix_to_virt(__end_of_fixed_addresses) <= MODULES_END);
157
158 cr4_init_shadow();
159
160 /* Kill off the identity-map trampoline */
161 reset_early_page_tables();
162
163 /* clear bss before set_intr_gate with early_idt_handler */
164 clear_bss();
165
166 for (i = 0; i < NUM_EXCEPTION_VECTORS; i++)
167 set_intr_gate(i, early_idt_handler_array[i]);
168 load_idt((const struct desc_ptr *)&idt_descr);
169
170 copy_bootdata(__va(real_mode_data));
171
172 /*
173 * Load microcode early on BSP.
174 */
175 load_ucode_bsp();
176
177 if (console_loglevel >= CONSOLE_LOGLEVEL_DEBUG)
178 early_printk("Kernel alive\n");
179
180 clear_page(init_level4_pgt);
181 /* set init_level4_pgt kernel high mapping*/
182 init_level4_pgt[511] = early_level4_pgt[511];
183
184 x86_64_start_reservations(real_mode_data);
185 }
186
x86_64_start_reservations(char * real_mode_data)187 void __init x86_64_start_reservations(char *real_mode_data)
188 {
189 /* version is always not zero if it is copied */
190 if (!boot_params.hdr.version)
191 copy_bootdata(__va(real_mode_data));
192
193 reserve_ebda_region();
194
195 start_kernel();
196 }
197