1
// SPDX-License-Identifier: GPL-2.0
2
/*
3
 * Copyright (C) 2000 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com)
4
 */
5

6
#include <linux/cpu.h>
7
#include <linux/delay.h>
8
#include <linux/init.h>
9
#include <linux/mm.h>
10
#include <linux/ctype.h>
11
#include <linux/module.h>
12
#include <linux/panic_notifier.h>
13
#include <linux/seq_file.h>
14
#include <linux/string.h>
15
#include <linux/string_choices.h>
16
#include <linux/utsname.h>
17
#include <linux/sched.h>
18
#include <linux/sched/task.h>
19
#include <linux/kmsg_dump.h>
20
#include <linux/suspend.h>
21
#include <linux/random.h>
22

23
#include <asm/processor.h>
24
#include <asm/cpufeature.h>
25
#include <asm/sections.h>
26
#include <asm/setup.h>
27
#include <asm/text-patching.h>
28
#include <as-layout.h>
29
#include <arch.h>
30
#include <init.h>
31
#include <kern.h>
32
#include <kern_util.h>
33
#include <mem_user.h>
34
#include <os.h>
35

36
#include "um_arch.h"
37

38
#define DEFAULT_COMMAND_LINE_ROOT "root=98:0"
39
#define DEFAULT_COMMAND_LINE_CONSOLE "console=tty0"
40

41
/* Changed in add_arg and setup_arch, which run before SMP is started */
42
static char __initdata command_line[COMMAND_LINE_SIZE] = { 0 };
43

44
static void __init add_arg(char *arg)
45
{
46
	if (strlen(command_line) + strlen(arg) + 1 > COMMAND_LINE_SIZE) {
47
		os_warn("add_arg: Too many command line arguments!\n");
48
		exit(1);
49
	}
50
	if (strlen(command_line) > 0)
51
		strcat(command_line, " ");
52
	strcat(command_line, arg);
53
}
54

55
/*
56
 * These fields are initialized at boot time and not changed.
57
 */
58
struct cpuinfo_um boot_cpu_data = {
59
	.loops_per_jiffy	= 0,
60
	.cache_alignment	= L1_CACHE_BYTES,
61
	.x86_capability		= { 0 }
62
};
63

64
EXPORT_SYMBOL(boot_cpu_data);
65

66

67
/* Changed in setup_arch, which is called in early boot */
68
static char host_info[(__NEW_UTS_LEN + 1) * 5];
69

70
static int show_cpuinfo(struct seq_file *m, void *v)
71
{
72
	int i = 0;
73

74
	seq_printf(m, "processor\t: %d\n", i);
75
	seq_printf(m, "vendor_id\t: User Mode Linux\n");
76
	seq_printf(m, "model name\t: UML\n");
77
	seq_printf(m, "mode\t\t: skas\n");
78
	seq_printf(m, "host\t\t: %s\n", host_info);
79
	seq_printf(m, "fpu\t\t: %s\n", str_yes_no(cpu_has(&boot_cpu_data, X86_FEATURE_FPU)));
80
	seq_printf(m, "flags\t\t:");
81
	for (i = 0; i < 32*NCAPINTS; i++)
82
		if (cpu_has(&boot_cpu_data, i) && (x86_cap_flags[i] != NULL))
83
			seq_printf(m, " %s", x86_cap_flags[i]);
84
	seq_printf(m, "\n");
85
	seq_printf(m, "cache_alignment\t: %d\n", boot_cpu_data.cache_alignment);
86
	seq_printf(m, "bogomips\t: %lu.%02lu\n",
87
		   loops_per_jiffy/(500000/HZ),
88
		   (loops_per_jiffy/(5000/HZ)) % 100);
89

90

91
	return 0;
92
}
93

94
static void *c_start(struct seq_file *m, loff_t *pos)
95
{
96
	return *pos < nr_cpu_ids ? &boot_cpu_data + *pos : NULL;
97
}
98

99
static void *c_next(struct seq_file *m, void *v, loff_t *pos)
100
{
101
	++*pos;
102
	return c_start(m, pos);
103
}
104

105
static void c_stop(struct seq_file *m, void *v)
106
{
107
}
108

109
const struct seq_operations cpuinfo_op = {
110
	.start	= c_start,
111
	.next	= c_next,
112
	.stop	= c_stop,
113
	.show	= show_cpuinfo,
114
};
115

116
/* Set in linux_main */
117
unsigned long uml_physmem;
118
EXPORT_SYMBOL(uml_physmem);
119

120
unsigned long uml_reserved; /* Also modified in mem_init */
121
unsigned long start_vm;
122
unsigned long end_vm;
123

124
/* Set in early boot */
125
static int have_root __initdata;
126
static int have_console __initdata;
127

128
/* Set in uml_mem_setup and modified in linux_main */
129
unsigned long long physmem_size = 64 * 1024 * 1024;
130
EXPORT_SYMBOL(physmem_size);
131

132
static const char *usage_string =
133
"User Mode Linux v%s\n"
134
"	available at http://user-mode-linux.sourceforge.net/\n\n";
135

136
static int __init uml_version_setup(char *line, int *add)
137
{
138
	/* Explicitly use printf() to show version in stdout */
139
	printf("%s\n", init_utsname()->release);
140
	exit(0);
141

142
	return 0;
143
}
144

145
__uml_setup("--version", uml_version_setup,
146
"--version\n"
147
"    Prints the version number of the kernel.\n\n"
148
);
149

150
static int __init uml_root_setup(char *line, int *add)
151
{
152
	have_root = 1;
153
	return 0;
154
}
155

156
__uml_setup("root=", uml_root_setup,
157
"root=<file containing the root fs>\n"
158
"    This is actually used by the generic kernel in exactly the same\n"
159
"    way as in any other kernel. If you configure a number of block\n"
160
"    devices and want to boot off something other than ubd0, you \n"
161
"    would use something like:\n"
162
"        root=/dev/ubd5\n\n"
163
);
164

165
static int __init uml_console_setup(char *line, int *add)
166
{
167
	have_console = 1;
168
	return 0;
169
}
170

171
__uml_setup("console=", uml_console_setup,
172
"console=<preferred console>\n"
173
"    Specify the preferred console output driver\n\n"
174
);
175

176
static int __init Usage(char *line, int *add)
177
{
178
	const char **p;
179

180
	printf(usage_string, init_utsname()->release);
181
	p = &__uml_help_start;
182
	/* Explicitly use printf() to show help in stdout */
183
	while (p < &__uml_help_end) {
184
		printf("%s", *p);
185
		p++;
186
	}
187
	exit(0);
188
	return 0;
189
}
190

191
__uml_setup("--help", Usage,
192
"--help\n"
193
"    Prints this message.\n\n"
194
);
195

196
static void __init uml_checksetup(char *line, int *add)
197
{
198
	struct uml_param *p;
199

200
	p = &__uml_setup_start;
201
	while (p < &__uml_setup_end) {
202
		size_t n;
203

204
		n = strlen(p->str);
205
		if (!strncmp(line, p->str, n) && p->setup_func(line + n, add))
206
			return;
207
		p++;
208
	}
209
}
210

211
static void __init uml_postsetup(void)
212
{
213
	initcall_t *p;
214

215
	p = &__uml_postsetup_start;
216
	while (p < &__uml_postsetup_end) {
217
		(*p)();
218
		p++;
219
	}
220
	return;
221
}
222

223
static int panic_exit(struct notifier_block *self, unsigned long unused1,
224
		      void *unused2)
225
{
226
	kmsg_dump(KMSG_DUMP_PANIC);
227
	bust_spinlocks(1);
228
	bust_spinlocks(0);
229
	uml_exitcode = 1;
230
	os_dump_core();
231

232
	return NOTIFY_DONE;
233
}
234

235
static struct notifier_block panic_exit_notifier = {
236
	.notifier_call	= panic_exit,
237
	.priority	= INT_MAX - 1, /* run as 2nd notifier, won't return */
238
};
239

240
void uml_finishsetup(void)
241
{
242
	cpu_tasks[0] = &init_task;
243

244
	atomic_notifier_chain_register(&panic_notifier_list,
245
				       &panic_exit_notifier);
246

247
	uml_postsetup();
248

249
	new_thread_handler();
250
}
251

252
/* Set during early boot */
253
unsigned long stub_start;
254
unsigned long task_size;
255
EXPORT_SYMBOL(task_size);
256

257
unsigned long host_task_size;
258

259
unsigned long brk_start;
260
unsigned long end_iomem;
261
EXPORT_SYMBOL(end_iomem);
262

263
#define MIN_VMALLOC (32 * 1024 * 1024)
264

265
static void __init parse_host_cpu_flags(char *line)
266
{
267
	int i;
268
	for (i = 0; i < 32*NCAPINTS; i++) {
269
		if ((x86_cap_flags[i] != NULL) && strstr(line, x86_cap_flags[i]))
270
			set_cpu_cap(&boot_cpu_data, i);
271
	}
272
}
273

274
static void __init parse_cache_line(char *line)
275
{
276
	long res;
277
	char *to_parse = strstr(line, ":");
278
	if (to_parse) {
279
		to_parse++;
280
		while (*to_parse != 0 && isspace(*to_parse)) {
281
			to_parse++;
282
		}
283
		if (kstrtoul(to_parse, 10, &res) == 0 && is_power_of_2(res))
284
			boot_cpu_data.cache_alignment = res;
285
		else
286
			boot_cpu_data.cache_alignment = L1_CACHE_BYTES;
287
	}
288
}
289

290
static unsigned long __init get_top_address(char **envp)
291
{
292
	unsigned long top_addr = (unsigned long) &top_addr;
293
	int i;
294

295
	/* The earliest variable should be after the program name in ELF */
296
	for (i = 0; envp[i]; i++) {
297
		if ((unsigned long) envp[i] > top_addr)
298
			top_addr = (unsigned long) envp[i];
299
	}
300

301
	top_addr &= ~(UM_KERN_PAGE_SIZE - 1);
302
	top_addr += UM_KERN_PAGE_SIZE;
303

304
	return top_addr;
305
}
306

307
int __init linux_main(int argc, char **argv, char **envp)
308
{
309
	unsigned long avail, diff;
310
	unsigned long virtmem_size, max_physmem;
311
	unsigned long stack;
312
	unsigned int i;
313
	int add;
314

315
	for (i = 1; i < argc; i++) {
316
		if ((i == 1) && (argv[i][0] == ' '))
317
			continue;
318
		add = 1;
319
		uml_checksetup(argv[i], &add);
320
		if (add)
321
			add_arg(argv[i]);
322
	}
323
	if (have_root == 0)
324
		add_arg(DEFAULT_COMMAND_LINE_ROOT);
325

326
	if (have_console == 0)
327
		add_arg(DEFAULT_COMMAND_LINE_CONSOLE);
328

329
	host_task_size = get_top_address(envp);
330
	/* reserve a few pages for the stubs */
331
	stub_start = host_task_size - STUB_SIZE;
332
	host_task_size = stub_start;
333

334
	/* Limit TASK_SIZE to what is addressable by the page table */
335
	task_size = host_task_size;
336
	if (task_size > (unsigned long long) PTRS_PER_PGD * PGDIR_SIZE)
337
		task_size = PTRS_PER_PGD * PGDIR_SIZE;
338

339
	/*
340
	 * TASK_SIZE needs to be PGDIR_SIZE aligned or else exit_mmap craps
341
	 * out
342
	 */
343
	task_size = task_size & PGDIR_MASK;
344

345
	/* OS sanity checks that need to happen before the kernel runs */
346
	os_early_checks();
347

348
	get_host_cpu_features(parse_host_cpu_flags, parse_cache_line);
349

350
	brk_start = (unsigned long) sbrk(0);
351

352
	/*
353
	 * Increase physical memory size for exec-shield users
354
	 * so they actually get what they asked for. This should
355
	 * add zero for non-exec shield users
356
	 */
357

358
	diff = UML_ROUND_UP(brk_start) - UML_ROUND_UP(&_end);
359
	if (diff > 1024 * 1024) {
360
		os_info("Adding %ld bytes to physical memory to account for "
361
			"exec-shield gap\n", diff);
362
		physmem_size += UML_ROUND_UP(brk_start) - UML_ROUND_UP(&_end);
363
	}
364

365
	uml_physmem = (unsigned long) __binary_start & PAGE_MASK;
366

367
	/* Reserve up to 4M after the current brk */
368
	uml_reserved = ROUND_4M(brk_start) + (1 << 22);
369

370
	setup_machinename(init_utsname()->machine);
371

372
	physmem_size = (physmem_size + PAGE_SIZE - 1) & PAGE_MASK;
373
	iomem_size = (iomem_size + PAGE_SIZE - 1) & PAGE_MASK;
374

375
	max_physmem = TASK_SIZE - uml_physmem - iomem_size - MIN_VMALLOC;
376
	if (physmem_size > max_physmem) {
377
		physmem_size = max_physmem;
378
		os_info("Physical memory size shrunk to %llu bytes\n",
379
			physmem_size);
380
	}
381

382
	high_physmem = uml_physmem + physmem_size;
383
	end_iomem = high_physmem + iomem_size;
384

385
	start_vm = VMALLOC_START;
386

387
	virtmem_size = physmem_size;
388
	stack = (unsigned long) argv;
389
	stack &= ~(1024 * 1024 - 1);
390
	avail = stack - start_vm;
391
	if (physmem_size > avail)
392
		virtmem_size = avail;
393
	end_vm = start_vm + virtmem_size;
394

395
	if (virtmem_size < physmem_size)
396
		os_info("Kernel virtual memory size shrunk to %lu bytes\n",
397
			virtmem_size);
398

399
	arch_task_struct_size = sizeof(struct task_struct) + host_fp_size;
400

401
	os_flush_stdout();
402

403
	return start_uml();
404
}
405

406
int __init __weak read_initrd(void)
407
{
408
	return 0;
409
}
410

411
void __init setup_arch(char **cmdline_p)
412
{
413
	u8 rng_seed[32];
414

415
	stack_protections((unsigned long) init_task.stack);
416
	setup_physmem(uml_physmem, uml_reserved, physmem_size);
417
	uml_dtb_init();
418
	read_initrd();
419

420
	paging_init();
421
	strscpy(boot_command_line, command_line, COMMAND_LINE_SIZE);
422
	*cmdline_p = command_line;
423
	setup_hostinfo(host_info, sizeof host_info);
424

425
	if (os_getrandom(rng_seed, sizeof(rng_seed), 0) == sizeof(rng_seed)) {
426
		add_bootloader_randomness(rng_seed, sizeof(rng_seed));
427
		memzero_explicit(rng_seed, sizeof(rng_seed));
428
	}
429
}
430

431
void __init arch_cpu_finalize_init(void)
432
{
433
	arch_check_bugs();
434
	os_check_bugs();
435
}
436

437
void apply_seal_endbr(s32 *start, s32 *end)
438
{
439
}
440

441
void apply_retpolines(s32 *start, s32 *end)
442
{
443
}
444

445
void apply_returns(s32 *start, s32 *end)
446
{
447
}
448

449
void apply_fineibt(s32 *start_retpoline, s32 *end_retpoline,
450
		   s32 *start_cfi, s32 *end_cfi)
451
{
452
}
453

454
void apply_alternatives(struct alt_instr *start, struct alt_instr *end)
455
{
456
}
457

458
void *text_poke(void *addr, const void *opcode, size_t len)
459
{
460
	/*
461
	 * In UML, the only reference to this function is in
462
	 * apply_relocate_add(), which shouldn't ever actually call this
463
	 * because UML doesn't have live patching.
464
	 */
465
	WARN_ON(1);
466

467
	return memcpy(addr, opcode, len);
468
}
469

470
void *text_poke_copy(void *addr, const void *opcode, size_t len)
471
{
472
	return text_poke(addr, opcode, len);
473
}
474

475
void smp_text_poke_sync_each_cpu(void)
476
{
477
}
478

479
void uml_pm_wake(void)
480
{
481
	pm_system_wakeup();
482
}
483

484
#ifdef CONFIG_PM_SLEEP
485
static int um_suspend_valid(suspend_state_t state)
486
{
487
	return state == PM_SUSPEND_MEM;
488
}
489

490
static int um_suspend_prepare(void)
491
{
492
	um_irqs_suspend();
493
	return 0;
494
}
495

496
static int um_suspend_enter(suspend_state_t state)
497
{
498
	if (WARN_ON(state != PM_SUSPEND_MEM))
499
		return -EINVAL;
500

501
	/*
502
	 * This is identical to the idle sleep, but we've just
503
	 * (during suspend) turned off all interrupt sources
504
	 * except for the ones we want, so now we can only wake
505
	 * up on something we actually want to wake up on. All
506
	 * timing has also been suspended.
507
	 */
508
	um_idle_sleep();
509
	return 0;
510
}
511

512
static void um_suspend_finish(void)
513
{
514
	um_irqs_resume();
515
}
516

517
const struct platform_suspend_ops um_suspend_ops = {
518
	.valid = um_suspend_valid,
519
	.prepare = um_suspend_prepare,
520
	.enter = um_suspend_enter,
521
	.finish = um_suspend_finish,
522
};
523

524
static int init_pm_wake_signal(void)
525
{
526
	/*
527
	 * In external time-travel mode we can't use signals to wake up
528
	 * since that would mess with the scheduling. We'll have to do
529
	 * some additional work to support wakeup on virtio devices or
530
	 * similar, perhaps implementing a fake RTC controller that can
531
	 * trigger wakeup (and request the appropriate scheduling from
532
	 * the external scheduler when going to suspend.)
533
	 */
534
	if (time_travel_mode != TT_MODE_EXTERNAL)
535
		register_pm_wake_signal();
536

537
	suspend_set_ops(&um_suspend_ops);
538

539
	return 0;
540
}
541

542
late_initcall(init_pm_wake_signal);
543
#endif
544

545