Loading...
1/*
2 * misc.c
3 *
4 * This is a collection of several routines from gzip-1.0.3
5 * adapted for Linux.
6 *
7 * malloc by Hannu Savolainen 1993 and Matthias Urlichs 1994
8 * puts by Nick Holloway 1993, better puts by Martin Mares 1995
9 * High loaded stuff by Hans Lermen & Werner Almesberger, Feb. 1996
10 */
11
12#include "misc.h"
13
14/* WARNING!!
15 * This code is compiled with -fPIC and it is relocated dynamically
16 * at run time, but no relocation processing is performed.
17 * This means that it is not safe to place pointers in static structures.
18 */
19
20/*
21 * Getting to provable safe in place decompression is hard.
22 * Worst case behaviours need to be analyzed.
23 * Background information:
24 *
25 * The file layout is:
26 * magic[2]
27 * method[1]
28 * flags[1]
29 * timestamp[4]
30 * extraflags[1]
31 * os[1]
32 * compressed data blocks[N]
33 * crc[4] orig_len[4]
34 *
35 * resulting in 18 bytes of non compressed data overhead.
36 *
37 * Files divided into blocks
38 * 1 bit (last block flag)
39 * 2 bits (block type)
40 *
41 * 1 block occurs every 32K -1 bytes or when there 50% compression
42 * has been achieved. The smallest block type encoding is always used.
43 *
44 * stored:
45 * 32 bits length in bytes.
46 *
47 * fixed:
48 * magic fixed tree.
49 * symbols.
50 *
51 * dynamic:
52 * dynamic tree encoding.
53 * symbols.
54 *
55 *
56 * The buffer for decompression in place is the length of the
57 * uncompressed data, plus a small amount extra to keep the algorithm safe.
58 * The compressed data is placed at the end of the buffer. The output
59 * pointer is placed at the start of the buffer and the input pointer
60 * is placed where the compressed data starts. Problems will occur
61 * when the output pointer overruns the input pointer.
62 *
63 * The output pointer can only overrun the input pointer if the input
64 * pointer is moving faster than the output pointer. A condition only
65 * triggered by data whose compressed form is larger than the uncompressed
66 * form.
67 *
68 * The worst case at the block level is a growth of the compressed data
69 * of 5 bytes per 32767 bytes.
70 *
71 * The worst case internal to a compressed block is very hard to figure.
72 * The worst case can at least be boundined by having one bit that represents
73 * 32764 bytes and then all of the rest of the bytes representing the very
74 * very last byte.
75 *
76 * All of which is enough to compute an amount of extra data that is required
77 * to be safe. To avoid problems at the block level allocating 5 extra bytes
78 * per 32767 bytes of data is sufficient. To avoind problems internal to a
79 * block adding an extra 32767 bytes (the worst case uncompressed block size)
80 * is sufficient, to ensure that in the worst case the decompressed data for
81 * block will stop the byte before the compressed data for a block begins.
82 * To avoid problems with the compressed data's meta information an extra 18
83 * bytes are needed. Leading to the formula:
84 *
85 * extra_bytes = (uncompressed_size >> 12) + 32768 + 18 + decompressor_size.
86 *
87 * Adding 8 bytes per 32K is a bit excessive but much easier to calculate.
88 * Adding 32768 instead of 32767 just makes for round numbers.
89 * Adding the decompressor_size is necessary as it musht live after all
90 * of the data as well. Last I measured the decompressor is about 14K.
91 * 10K of actual data and 4K of bss.
92 *
93 */
94
95/*
96 * gzip declarations
97 */
98#define STATIC static
99
100#undef memset
101#undef memcpy
102#define memzero(s, n) memset((s), 0, (n))
103
104
105static void error(char *m);
106
107/*
108 * This is set up by the setup-routine at boot-time
109 */
110struct boot_params *real_mode; /* Pointer to real-mode data */
111static int quiet;
112static int debug;
113
114void *memset(void *s, int c, size_t n);
115void *memcpy(void *dest, const void *src, size_t n);
116
117#ifdef CONFIG_X86_64
118#define memptr long
119#else
120#define memptr unsigned
121#endif
122
123static memptr free_mem_ptr;
124static memptr free_mem_end_ptr;
125
126static char *vidmem;
127static int vidport;
128static int lines, cols;
129
130#ifdef CONFIG_KERNEL_GZIP
131#include "../../../../lib/decompress_inflate.c"
132#endif
133
134#ifdef CONFIG_KERNEL_BZIP2
135#include "../../../../lib/decompress_bunzip2.c"
136#endif
137
138#ifdef CONFIG_KERNEL_LZMA
139#include "../../../../lib/decompress_unlzma.c"
140#endif
141
142#ifdef CONFIG_KERNEL_XZ
143#include "../../../../lib/decompress_unxz.c"
144#endif
145
146#ifdef CONFIG_KERNEL_LZO
147#include "../../../../lib/decompress_unlzo.c"
148#endif
149
150static void scroll(void)
151{
152 int i;
153
154 memcpy(vidmem, vidmem + cols * 2, (lines - 1) * cols * 2);
155 for (i = (lines - 1) * cols * 2; i < lines * cols * 2; i += 2)
156 vidmem[i] = ' ';
157}
158
159#define XMTRDY 0x20
160
161#define TXR 0 /* Transmit register (WRITE) */
162#define LSR 5 /* Line Status */
163static void serial_putchar(int ch)
164{
165 unsigned timeout = 0xffff;
166
167 while ((inb(early_serial_base + LSR) & XMTRDY) == 0 && --timeout)
168 cpu_relax();
169
170 outb(ch, early_serial_base + TXR);
171}
172
173void __putstr(int error, const char *s)
174{
175 int x, y, pos;
176 char c;
177
178#ifndef CONFIG_X86_VERBOSE_BOOTUP
179 if (!error)
180 return;
181#endif
182 if (early_serial_base) {
183 const char *str = s;
184 while (*str) {
185 if (*str == '\n')
186 serial_putchar('\r');
187 serial_putchar(*str++);
188 }
189 }
190
191 if (real_mode->screen_info.orig_video_mode == 0 &&
192 lines == 0 && cols == 0)
193 return;
194
195 x = real_mode->screen_info.orig_x;
196 y = real_mode->screen_info.orig_y;
197
198 while ((c = *s++) != '\0') {
199 if (c == '\n') {
200 x = 0;
201 if (++y >= lines) {
202 scroll();
203 y--;
204 }
205 } else {
206 vidmem[(x + cols * y) * 2] = c;
207 if (++x >= cols) {
208 x = 0;
209 if (++y >= lines) {
210 scroll();
211 y--;
212 }
213 }
214 }
215 }
216
217 real_mode->screen_info.orig_x = x;
218 real_mode->screen_info.orig_y = y;
219
220 pos = (x + cols * y) * 2; /* Update cursor position */
221 outb(14, vidport);
222 outb(0xff & (pos >> 9), vidport+1);
223 outb(15, vidport);
224 outb(0xff & (pos >> 1), vidport+1);
225}
226
227void *memset(void *s, int c, size_t n)
228{
229 int i;
230 char *ss = s;
231
232 for (i = 0; i < n; i++)
233 ss[i] = c;
234 return s;
235}
236#ifdef CONFIG_X86_32
237void *memcpy(void *dest, const void *src, size_t n)
238{
239 int d0, d1, d2;
240 asm volatile(
241 "rep ; movsl\n\t"
242 "movl %4,%%ecx\n\t"
243 "rep ; movsb\n\t"
244 : "=&c" (d0), "=&D" (d1), "=&S" (d2)
245 : "0" (n >> 2), "g" (n & 3), "1" (dest), "2" (src)
246 : "memory");
247
248 return dest;
249}
250#else
251void *memcpy(void *dest, const void *src, size_t n)
252{
253 long d0, d1, d2;
254 asm volatile(
255 "rep ; movsq\n\t"
256 "movq %4,%%rcx\n\t"
257 "rep ; movsb\n\t"
258 : "=&c" (d0), "=&D" (d1), "=&S" (d2)
259 : "0" (n >> 3), "g" (n & 7), "1" (dest), "2" (src)
260 : "memory");
261
262 return dest;
263}
264#endif
265
266static void error(char *x)
267{
268 __putstr(1, "\n\n");
269 __putstr(1, x);
270 __putstr(1, "\n\n -- System halted");
271
272 while (1)
273 asm("hlt");
274}
275
276static void parse_elf(void *output)
277{
278#ifdef CONFIG_X86_64
279 Elf64_Ehdr ehdr;
280 Elf64_Phdr *phdrs, *phdr;
281#else
282 Elf32_Ehdr ehdr;
283 Elf32_Phdr *phdrs, *phdr;
284#endif
285 void *dest;
286 int i;
287
288 memcpy(&ehdr, output, sizeof(ehdr));
289 if (ehdr.e_ident[EI_MAG0] != ELFMAG0 ||
290 ehdr.e_ident[EI_MAG1] != ELFMAG1 ||
291 ehdr.e_ident[EI_MAG2] != ELFMAG2 ||
292 ehdr.e_ident[EI_MAG3] != ELFMAG3) {
293 error("Kernel is not a valid ELF file");
294 return;
295 }
296
297 if (!quiet)
298 putstr("Parsing ELF... ");
299
300 phdrs = malloc(sizeof(*phdrs) * ehdr.e_phnum);
301 if (!phdrs)
302 error("Failed to allocate space for phdrs");
303
304 memcpy(phdrs, output + ehdr.e_phoff, sizeof(*phdrs) * ehdr.e_phnum);
305
306 for (i = 0; i < ehdr.e_phnum; i++) {
307 phdr = &phdrs[i];
308
309 switch (phdr->p_type) {
310 case PT_LOAD:
311#ifdef CONFIG_RELOCATABLE
312 dest = output;
313 dest += (phdr->p_paddr - LOAD_PHYSICAL_ADDR);
314#else
315 dest = (void *)(phdr->p_paddr);
316#endif
317 memcpy(dest,
318 output + phdr->p_offset,
319 phdr->p_filesz);
320 break;
321 default: /* Ignore other PT_* */ break;
322 }
323 }
324}
325
326asmlinkage void decompress_kernel(void *rmode, memptr heap,
327 unsigned char *input_data,
328 unsigned long input_len,
329 unsigned char *output)
330{
331 real_mode = rmode;
332
333 if (cmdline_find_option_bool("quiet"))
334 quiet = 1;
335 if (cmdline_find_option_bool("debug"))
336 debug = 1;
337
338 if (real_mode->screen_info.orig_video_mode == 7) {
339 vidmem = (char *) 0xb0000;
340 vidport = 0x3b4;
341 } else {
342 vidmem = (char *) 0xb8000;
343 vidport = 0x3d4;
344 }
345
346 lines = real_mode->screen_info.orig_video_lines;
347 cols = real_mode->screen_info.orig_video_cols;
348
349 console_init();
350 if (debug)
351 putstr("early console in decompress_kernel\n");
352
353 free_mem_ptr = heap; /* Heap */
354 free_mem_end_ptr = heap + BOOT_HEAP_SIZE;
355
356 if ((unsigned long)output & (MIN_KERNEL_ALIGN - 1))
357 error("Destination address inappropriately aligned");
358#ifdef CONFIG_X86_64
359 if (heap > 0x3fffffffffffUL)
360 error("Destination address too large");
361#else
362 if (heap > ((-__PAGE_OFFSET-(128<<20)-1) & 0x7fffffff))
363 error("Destination address too large");
364#endif
365#ifndef CONFIG_RELOCATABLE
366 if ((unsigned long)output != LOAD_PHYSICAL_ADDR)
367 error("Wrong destination address");
368#endif
369
370 if (!quiet)
371 putstr("\nDecompressing Linux... ");
372 decompress(input_data, input_len, NULL, NULL, output, NULL, error);
373 parse_elf(output);
374 if (!quiet)
375 putstr("done.\nBooting the kernel.\n");
376 return;
377}
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * misc.c
4 *
5 * This is a collection of several routines used to extract the kernel
6 * which includes KASLR relocation, decompression, ELF parsing, and
7 * relocation processing. Additionally included are the screen and serial
8 * output functions and related debugging support functions.
9 *
10 * malloc by Hannu Savolainen 1993 and Matthias Urlichs 1994
11 * puts by Nick Holloway 1993, better puts by Martin Mares 1995
12 * High loaded stuff by Hans Lermen & Werner Almesberger, Feb. 1996
13 */
14
15#include "misc.h"
16#include "error.h"
17#include "pgtable.h"
18#include "../string.h"
19#include "../voffset.h"
20#include <asm/bootparam_utils.h>
21
22/*
23 * WARNING!!
24 * This code is compiled with -fPIC and it is relocated dynamically at
25 * run time, but no relocation processing is performed. This means that
26 * it is not safe to place pointers in static structures.
27 */
28
29/* Macros used by the included decompressor code below. */
30#define STATIC static
31/* Define an externally visible malloc()/free(). */
32#define MALLOC_VISIBLE
33#include <linux/decompress/mm.h>
34
35/*
36 * Provide definitions of memzero and memmove as some of the decompressors will
37 * try to define their own functions if these are not defined as macros.
38 */
39#define memzero(s, n) memset((s), 0, (n))
40#ifndef memmove
41#define memmove memmove
42/* Functions used by the included decompressor code below. */
43void *memmove(void *dest, const void *src, size_t n);
44#endif
45
46/*
47 * This is set up by the setup-routine at boot-time
48 */
49struct boot_params *boot_params_ptr;
50
51struct port_io_ops pio_ops;
52
53memptr free_mem_ptr;
54memptr free_mem_end_ptr;
55int spurious_nmi_count;
56
57static char *vidmem;
58static int vidport;
59
60/* These might be accessed before .bss is cleared, so use .data instead. */
61static int lines __section(".data");
62static int cols __section(".data");
63
64#ifdef CONFIG_KERNEL_GZIP
65#include "../../../../lib/decompress_inflate.c"
66#endif
67
68#ifdef CONFIG_KERNEL_BZIP2
69#include "../../../../lib/decompress_bunzip2.c"
70#endif
71
72#ifdef CONFIG_KERNEL_LZMA
73#include "../../../../lib/decompress_unlzma.c"
74#endif
75
76#ifdef CONFIG_KERNEL_XZ
77#include "../../../../lib/decompress_unxz.c"
78#endif
79
80#ifdef CONFIG_KERNEL_LZO
81#include "../../../../lib/decompress_unlzo.c"
82#endif
83
84#ifdef CONFIG_KERNEL_LZ4
85#include "../../../../lib/decompress_unlz4.c"
86#endif
87
88#ifdef CONFIG_KERNEL_ZSTD
89#include "../../../../lib/decompress_unzstd.c"
90#endif
91/*
92 * NOTE: When adding a new decompressor, please update the analysis in
93 * ../header.S.
94 */
95
96static void scroll(void)
97{
98 int i;
99
100 memmove(vidmem, vidmem + cols * 2, (lines - 1) * cols * 2);
101 for (i = (lines - 1) * cols * 2; i < lines * cols * 2; i += 2)
102 vidmem[i] = ' ';
103}
104
105#define XMTRDY 0x20
106
107#define TXR 0 /* Transmit register (WRITE) */
108#define LSR 5 /* Line Status */
109static void serial_putchar(int ch)
110{
111 unsigned timeout = 0xffff;
112
113 while ((inb(early_serial_base + LSR) & XMTRDY) == 0 && --timeout)
114 cpu_relax();
115
116 outb(ch, early_serial_base + TXR);
117}
118
119void __putstr(const char *s)
120{
121 int x, y, pos;
122 char c;
123
124 if (early_serial_base) {
125 const char *str = s;
126 while (*str) {
127 if (*str == '\n')
128 serial_putchar('\r');
129 serial_putchar(*str++);
130 }
131 }
132
133 if (lines == 0 || cols == 0)
134 return;
135
136 x = boot_params_ptr->screen_info.orig_x;
137 y = boot_params_ptr->screen_info.orig_y;
138
139 while ((c = *s++) != '\0') {
140 if (c == '\n') {
141 x = 0;
142 if (++y >= lines) {
143 scroll();
144 y--;
145 }
146 } else {
147 vidmem[(x + cols * y) * 2] = c;
148 if (++x >= cols) {
149 x = 0;
150 if (++y >= lines) {
151 scroll();
152 y--;
153 }
154 }
155 }
156 }
157
158 boot_params_ptr->screen_info.orig_x = x;
159 boot_params_ptr->screen_info.orig_y = y;
160
161 pos = (x + cols * y) * 2; /* Update cursor position */
162 outb(14, vidport);
163 outb(0xff & (pos >> 9), vidport+1);
164 outb(15, vidport);
165 outb(0xff & (pos >> 1), vidport+1);
166}
167
168static noinline void __putnum(unsigned long value, unsigned int base,
169 int mindig)
170{
171 char buf[8*sizeof(value)+1];
172 char *p;
173
174 p = buf + sizeof(buf);
175 *--p = '\0';
176
177 while (mindig-- > 0 || value) {
178 unsigned char digit = value % base;
179 digit += (digit >= 10) ? ('a'-10) : '0';
180 *--p = digit;
181
182 value /= base;
183 }
184
185 __putstr(p);
186}
187
188void __puthex(unsigned long value)
189{
190 __putnum(value, 16, sizeof(value)*2);
191}
192
193void __putdec(unsigned long value)
194{
195 __putnum(value, 10, 1);
196}
197
198#ifdef CONFIG_X86_NEED_RELOCS
199static void handle_relocations(void *output, unsigned long output_len,
200 unsigned long virt_addr)
201{
202 int *reloc;
203 unsigned long delta, map, ptr;
204 unsigned long min_addr = (unsigned long)output;
205 unsigned long max_addr = min_addr + (VO___bss_start - VO__text);
206
207 /*
208 * Calculate the delta between where vmlinux was linked to load
209 * and where it was actually loaded.
210 */
211 delta = min_addr - LOAD_PHYSICAL_ADDR;
212
213 /*
214 * The kernel contains a table of relocation addresses. Those
215 * addresses have the final load address of the kernel in virtual
216 * memory. We are currently working in the self map. So we need to
217 * create an adjustment for kernel memory addresses to the self map.
218 * This will involve subtracting out the base address of the kernel.
219 */
220 map = delta - __START_KERNEL_map;
221
222 /*
223 * 32-bit always performs relocations. 64-bit relocations are only
224 * needed if KASLR has chosen a different starting address offset
225 * from __START_KERNEL_map.
226 */
227 if (IS_ENABLED(CONFIG_X86_64))
228 delta = virt_addr - LOAD_PHYSICAL_ADDR;
229
230 if (!delta) {
231 debug_putstr("No relocation needed... ");
232 return;
233 }
234 debug_putstr("Performing relocations... ");
235
236 /*
237 * Process relocations: 32 bit relocations first then 64 bit after.
238 * Three sets of binary relocations are added to the end of the kernel
239 * before compression. Each relocation table entry is the kernel
240 * address of the location which needs to be updated stored as a
241 * 32-bit value which is sign extended to 64 bits.
242 *
243 * Format is:
244 *
245 * kernel bits...
246 * 0 - zero terminator for 64 bit relocations
247 * 64 bit relocation repeated
248 * 0 - zero terminator for inverse 32 bit relocations
249 * 32 bit inverse relocation repeated
250 * 0 - zero terminator for 32 bit relocations
251 * 32 bit relocation repeated
252 *
253 * So we work backwards from the end of the decompressed image.
254 */
255 for (reloc = output + output_len - sizeof(*reloc); *reloc; reloc--) {
256 long extended = *reloc;
257 extended += map;
258
259 ptr = (unsigned long)extended;
260 if (ptr < min_addr || ptr > max_addr)
261 error("32-bit relocation outside of kernel!\n");
262
263 *(uint32_t *)ptr += delta;
264 }
265#ifdef CONFIG_X86_64
266 while (*--reloc) {
267 long extended = *reloc;
268 extended += map;
269
270 ptr = (unsigned long)extended;
271 if (ptr < min_addr || ptr > max_addr)
272 error("inverse 32-bit relocation outside of kernel!\n");
273
274 *(int32_t *)ptr -= delta;
275 }
276 for (reloc--; *reloc; reloc--) {
277 long extended = *reloc;
278 extended += map;
279
280 ptr = (unsigned long)extended;
281 if (ptr < min_addr || ptr > max_addr)
282 error("64-bit relocation outside of kernel!\n");
283
284 *(uint64_t *)ptr += delta;
285 }
286#endif
287}
288#else
289static inline void handle_relocations(void *output, unsigned long output_len,
290 unsigned long virt_addr)
291{ }
292#endif
293
294static size_t parse_elf(void *output)
295{
296#ifdef CONFIG_X86_64
297 Elf64_Ehdr ehdr;
298 Elf64_Phdr *phdrs, *phdr;
299#else
300 Elf32_Ehdr ehdr;
301 Elf32_Phdr *phdrs, *phdr;
302#endif
303 void *dest;
304 int i;
305
306 memcpy(&ehdr, output, sizeof(ehdr));
307 if (ehdr.e_ident[EI_MAG0] != ELFMAG0 ||
308 ehdr.e_ident[EI_MAG1] != ELFMAG1 ||
309 ehdr.e_ident[EI_MAG2] != ELFMAG2 ||
310 ehdr.e_ident[EI_MAG3] != ELFMAG3)
311 error("Kernel is not a valid ELF file");
312
313 debug_putstr("Parsing ELF... ");
314
315 phdrs = malloc(sizeof(*phdrs) * ehdr.e_phnum);
316 if (!phdrs)
317 error("Failed to allocate space for phdrs");
318
319 memcpy(phdrs, output + ehdr.e_phoff, sizeof(*phdrs) * ehdr.e_phnum);
320
321 for (i = 0; i < ehdr.e_phnum; i++) {
322 phdr = &phdrs[i];
323
324 switch (phdr->p_type) {
325 case PT_LOAD:
326#ifdef CONFIG_X86_64
327 if ((phdr->p_align % 0x200000) != 0)
328 error("Alignment of LOAD segment isn't multiple of 2MB");
329#endif
330#ifdef CONFIG_RELOCATABLE
331 dest = output;
332 dest += (phdr->p_paddr - LOAD_PHYSICAL_ADDR);
333#else
334 dest = (void *)(phdr->p_paddr);
335#endif
336 memmove(dest, output + phdr->p_offset, phdr->p_filesz);
337 break;
338 default: /* Ignore other PT_* */ break;
339 }
340 }
341
342 free(phdrs);
343
344 return ehdr.e_entry - LOAD_PHYSICAL_ADDR;
345}
346
347const unsigned long kernel_text_size = VO___start_rodata - VO__text;
348const unsigned long kernel_total_size = VO__end - VO__text;
349
350static u8 boot_heap[BOOT_HEAP_SIZE] __aligned(4);
351
352extern unsigned char input_data[];
353extern unsigned int input_len, output_len;
354
355unsigned long decompress_kernel(unsigned char *outbuf, unsigned long virt_addr,
356 void (*error)(char *x))
357{
358 unsigned long entry;
359
360 if (!free_mem_ptr) {
361 free_mem_ptr = (unsigned long)boot_heap;
362 free_mem_end_ptr = (unsigned long)boot_heap + sizeof(boot_heap);
363 }
364
365 if (__decompress(input_data, input_len, NULL, NULL, outbuf, output_len,
366 NULL, error) < 0)
367 return ULONG_MAX;
368
369 entry = parse_elf(outbuf);
370 handle_relocations(outbuf, output_len, virt_addr);
371
372 return entry;
373}
374
375/*
376 * Set the memory encryption xloadflag based on the mem_encrypt= command line
377 * parameter, if provided.
378 */
379static void parse_mem_encrypt(struct setup_header *hdr)
380{
381 int on = cmdline_find_option_bool("mem_encrypt=on");
382 int off = cmdline_find_option_bool("mem_encrypt=off");
383
384 if (on > off)
385 hdr->xloadflags |= XLF_MEM_ENCRYPTION;
386}
387
388static void early_sev_detect(void)
389{
390 /*
391 * Accessing video memory causes guest termination because
392 * the boot stage2 #VC handler of SEV-ES/SNP guests does not
393 * support MMIO handling and kexec -c adds screen_info to the
394 * boot parameters passed to the kexec kernel, which causes
395 * console output to be dumped to both video and serial.
396 */
397 if (sev_status & MSR_AMD64_SEV_ES_ENABLED)
398 lines = cols = 0;
399}
400
401/*
402 * The compressed kernel image (ZO), has been moved so that its position
403 * is against the end of the buffer used to hold the uncompressed kernel
404 * image (VO) and the execution environment (.bss, .brk), which makes sure
405 * there is room to do the in-place decompression. (See header.S for the
406 * calculations.)
407 *
408 * |-----compressed kernel image------|
409 * V V
410 * 0 extract_offset +INIT_SIZE
411 * |-----------|---------------|-------------------------|--------|
412 * | | | |
413 * VO__text startup_32 of ZO VO__end ZO__end
414 * ^ ^
415 * |-------uncompressed kernel image---------|
416 *
417 */
418asmlinkage __visible void *extract_kernel(void *rmode, unsigned char *output)
419{
420 unsigned long virt_addr = LOAD_PHYSICAL_ADDR;
421 memptr heap = (memptr)boot_heap;
422 unsigned long needed_size;
423 size_t entry_offset;
424
425 /* Retain x86 boot parameters pointer passed from startup_32/64. */
426 boot_params_ptr = rmode;
427
428 /* Clear flags intended for solely in-kernel use. */
429 boot_params_ptr->hdr.loadflags &= ~KASLR_FLAG;
430
431 parse_mem_encrypt(&boot_params_ptr->hdr);
432
433 sanitize_boot_params(boot_params_ptr);
434
435 if (boot_params_ptr->screen_info.orig_video_mode == 7) {
436 vidmem = (char *) 0xb0000;
437 vidport = 0x3b4;
438 } else {
439 vidmem = (char *) 0xb8000;
440 vidport = 0x3d4;
441 }
442
443 lines = boot_params_ptr->screen_info.orig_video_lines;
444 cols = boot_params_ptr->screen_info.orig_video_cols;
445
446 init_default_io_ops();
447
448 /*
449 * Detect TDX guest environment.
450 *
451 * It has to be done before console_init() in order to use
452 * paravirtualized port I/O operations if needed.
453 */
454 early_tdx_detect();
455
456 early_sev_detect();
457
458 console_init();
459
460 /*
461 * Save RSDP address for later use. Have this after console_init()
462 * so that early debugging output from the RSDP parsing code can be
463 * collected.
464 */
465 boot_params_ptr->acpi_rsdp_addr = get_rsdp_addr();
466
467 debug_putstr("early console in extract_kernel\n");
468
469 free_mem_ptr = heap; /* Heap */
470 free_mem_end_ptr = heap + BOOT_HEAP_SIZE;
471
472 /*
473 * The memory hole needed for the kernel is the larger of either
474 * the entire decompressed kernel plus relocation table, or the
475 * entire decompressed kernel plus .bss and .brk sections.
476 *
477 * On X86_64, the memory is mapped with PMD pages. Round the
478 * size up so that the full extent of PMD pages mapped is
479 * included in the check against the valid memory table
480 * entries. This ensures the full mapped area is usable RAM
481 * and doesn't include any reserved areas.
482 */
483 needed_size = max_t(unsigned long, output_len, kernel_total_size);
484#ifdef CONFIG_X86_64
485 needed_size = ALIGN(needed_size, MIN_KERNEL_ALIGN);
486#endif
487
488 /* Report initial kernel position details. */
489 debug_putaddr(input_data);
490 debug_putaddr(input_len);
491 debug_putaddr(output);
492 debug_putaddr(output_len);
493 debug_putaddr(kernel_total_size);
494 debug_putaddr(needed_size);
495
496#ifdef CONFIG_X86_64
497 /* Report address of 32-bit trampoline */
498 debug_putaddr(trampoline_32bit);
499#endif
500
501 choose_random_location((unsigned long)input_data, input_len,
502 (unsigned long *)&output,
503 needed_size,
504 &virt_addr);
505
506 /* Validate memory location choices. */
507 if ((unsigned long)output & (MIN_KERNEL_ALIGN - 1))
508 error("Destination physical address inappropriately aligned");
509 if (virt_addr & (MIN_KERNEL_ALIGN - 1))
510 error("Destination virtual address inappropriately aligned");
511#ifdef CONFIG_X86_64
512 if (heap > 0x3fffffffffffUL)
513 error("Destination address too large");
514 if (virt_addr + needed_size > KERNEL_IMAGE_SIZE)
515 error("Destination virtual address is beyond the kernel mapping area");
516#else
517 if (heap > ((-__PAGE_OFFSET-(128<<20)-1) & 0x7fffffff))
518 error("Destination address too large");
519#endif
520#ifndef CONFIG_RELOCATABLE
521 if (virt_addr != LOAD_PHYSICAL_ADDR)
522 error("Destination virtual address changed when not relocatable");
523#endif
524
525 debug_putstr("\nDecompressing Linux... ");
526
527 if (init_unaccepted_memory()) {
528 debug_putstr("Accepting memory... ");
529 accept_memory(__pa(output), needed_size);
530 }
531
532 entry_offset = decompress_kernel(output, virt_addr, error);
533
534 debug_putstr("done.\nBooting the kernel (entry_offset: 0x");
535 debug_puthex(entry_offset);
536 debug_putstr(").\n");
537
538 /* Disable exception handling before booting the kernel */
539 cleanup_exception_handling();
540
541 if (spurious_nmi_count) {
542 error_putstr("Spurious early NMIs ignored: ");
543 error_putdec(spurious_nmi_count);
544 error_putstr("\n");
545 }
546
547 return output + entry_offset;
548}