1 #include <linux/efi.h>
2 #include <asm/e820/types.h>
3 #include <asm/processor.h>
4 #include <asm/efi.h>
5 #include "pgtable.h"
6 #include "../string.h"
7
8 /*
9 * __force_order is used by special_insns.h asm code to force instruction
10 * serialization.
11 *
12 * It is not referenced from the code, but GCC < 5 with -fPIE would fail
13 * due to an undefined symbol. Define it to make these ancient GCCs work.
14 */
15 unsigned long __force_order;
16
17 #define BIOS_START_MIN 0x20000U /* 128K, less than this is insane */
18 #define BIOS_START_MAX 0x9f000U /* 640K, absolute maximum */
19
20 struct paging_config {
21 unsigned long trampoline_start;
22 unsigned long l5_required;
23 };
24
25 /* Buffer to preserve trampoline memory */
26 static char trampoline_save[TRAMPOLINE_32BIT_SIZE];
27
28 /*
29 * Trampoline address will be printed by extract_kernel() for debugging
30 * purposes.
31 *
32 * Avoid putting the pointer into .bss as it will be cleared between
33 * paging_prepare() and extract_kernel().
34 */
35 unsigned long *trampoline_32bit __section(.data);
36
37 extern struct boot_params *boot_params;
38 int cmdline_find_option_bool(const char *option);
39
40 static unsigned long find_trampoline_placement(void)
41 {
42 unsigned long bios_start = 0, ebda_start = 0;
43 struct boot_e820_entry *entry;
44 char *signature;
45 int i;
46
47 /*
48 * Find a suitable spot for the trampoline.
49 * This code is based on reserve_bios_regions().
50 */
51
52 /*
53 * EFI systems may not provide legacy ROM. The memory may not be mapped
54 * at all.
55 *
56 * Only look for values in the legacy ROM for non-EFI system.
57 */
58 signature = (char *)&boot_params->efi_info.efi_loader_signature;
59 if (strncmp(signature, EFI32_LOADER_SIGNATURE, 4) &&
60 strncmp(signature, EFI64_LOADER_SIGNATURE, 4)) {
61 ebda_start = *(unsigned short *)0x40e << 4;
62 bios_start = *(unsigned short *)0x413 << 10;
63 }
64
65 if (bios_start < BIOS_START_MIN || bios_start > BIOS_START_MAX)
66 bios_start = BIOS_START_MAX;
67
68 if (ebda_start > BIOS_START_MIN && ebda_start < bios_start)
69 bios_start = ebda_start;
70
71 bios_start = round_down(bios_start, PAGE_SIZE);
72
73 /* Find the first usable memory region under bios_start. */
74 for (i = boot_params->e820_entries - 1; i >= 0; i--) {
75 unsigned long new = bios_start;
76
77 entry = &boot_params->e820_table[i];
78
79 /* Skip all entries above bios_start. */
80 if (bios_start <= entry->addr)
81 continue;
82
83 /* Skip non-RAM entries. */
84 if (entry->type != E820_TYPE_RAM)
85 continue;
86
87 /* Adjust bios_start to the end of the entry if needed. */
88 if (bios_start > entry->addr + entry->size)
89 new = entry->addr + entry->size;
90
91 /* Keep bios_start page-aligned. */
92 new = round_down(new, PAGE_SIZE);
93
94 /* Skip the entry if it's too small. */
95 if (new - TRAMPOLINE_32BIT_SIZE < entry->addr)
96 continue;
97
98 /* Protect against underflow. */
99 if (new - TRAMPOLINE_32BIT_SIZE > bios_start)
100 break;
101
102 bios_start = new;
103 break;
104 }
105
106 /* Place the trampoline just below the end of low memory */
107 return bios_start - TRAMPOLINE_32BIT_SIZE;
108 }
109
110 struct paging_config paging_prepare(void *rmode)
111 {
112 struct paging_config paging_config = {};
113
114 /* Initialize boot_params. Required for cmdline_find_option_bool(). */
115 boot_params = rmode;
116
117 /*
118 * Check if LA57 is desired and supported.
119 *
120 * There are several parts to the check:
121 * - if the kernel supports 5-level paging: CONFIG_X86_5LEVEL=y
122 * - if user asked to disable 5-level paging: no5lvl in cmdline
123 * - if the machine supports 5-level paging:
124 * + CPUID leaf 7 is supported
125 * + the leaf has the feature bit set
126 *
127 * That's substitute for boot_cpu_has() in early boot code.
128 */
129 if (IS_ENABLED(CONFIG_X86_5LEVEL) &&
130 !cmdline_find_option_bool("no5lvl") &&
131 native_cpuid_eax(0) >= 7 &&
132 (native_cpuid_ecx(7) & (1 << (X86_FEATURE_LA57 & 31)))) {
133 paging_config.l5_required = 1;
134 }
135
136 paging_config.trampoline_start = find_trampoline_placement();
137
138 trampoline_32bit = (unsigned long *)paging_config.trampoline_start;
139
140 /* Preserve trampoline memory */
141 memcpy(trampoline_save, trampoline_32bit, TRAMPOLINE_32BIT_SIZE);
142
143 /* Clear trampoline memory first */
144 memset(trampoline_32bit, 0, TRAMPOLINE_32BIT_SIZE);
145
146 /* Copy trampoline code in place */
147 memcpy(trampoline_32bit + TRAMPOLINE_32BIT_CODE_OFFSET / sizeof(unsigned long),
148 &trampoline_32bit_src, TRAMPOLINE_32BIT_CODE_SIZE);
149
150 /*
151 * The code below prepares page table in trampoline memory.
152 *
153 * The new page table will be used by trampoline code for switching
154 * from 4- to 5-level paging or vice versa.
155 *
156 * If switching is not required, the page table is unused: trampoline
157 * code wouldn't touch CR3.
158 */
159
160 /*
161 * We are not going to use the page table in trampoline memory if we
162 * are already in the desired paging mode.
163 */
164 if (paging_config.l5_required == !!(native_read_cr4() & X86_CR4_LA57))
165 goto out;
166
167 if (paging_config.l5_required) {
168 /*
169 * For 4- to 5-level paging transition, set up current CR3 as
170 * the first and the only entry in a new top-level page table.
171 */
172 trampoline_32bit[TRAMPOLINE_32BIT_PGTABLE_OFFSET] = __native_read_cr3() | _PAGE_TABLE_NOENC;
173 } else {
174 unsigned long src;
175
176 /*
177 * For 5- to 4-level paging transition, copy page table pointed
178 * by first entry in the current top-level page table as our
179 * new top-level page table.
180 *
181 * We cannot just point to the page table from trampoline as it
182 * may be above 4G.
183 */
184 src = *(unsigned long *)__native_read_cr3() & PAGE_MASK;
185 memcpy(trampoline_32bit + TRAMPOLINE_32BIT_PGTABLE_OFFSET / sizeof(unsigned long),
186 (void *)src, PAGE_SIZE);
187 }
188
189 out:
190 return paging_config;
191 }
192
193 void cleanup_trampoline(void *pgtable)
194 {
195 void *trampoline_pgtable;
196
197 trampoline_pgtable = trampoline_32bit + TRAMPOLINE_32BIT_PGTABLE_OFFSET / sizeof(unsigned long);
198
199 /*
200 * Move the top level page table out of trampoline memory,
201 * if it's there.
202 */
203 if ((void *)__native_read_cr3() == trampoline_pgtable) {
204 memcpy(pgtable, trampoline_pgtable, PAGE_SIZE);
205 native_write_cr3((unsigned long)pgtable);
206 }
207
208 /* Restore trampoline memory */
209 memcpy(trampoline_32bit, trampoline_save, TRAMPOLINE_32BIT_SIZE);
210 }