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  * XXX This structure is used only in the non-SMP case.  Maybe this
58  * should be moved to smp.c.
59  */
60 struct cpuinfo_um boot_cpu_data = {
61 	.loops_per_jiffy	= 0,
62 	.ipi_pipe		= { -1, -1 },
63 	.cache_alignment	= L1_CACHE_BYTES,
64 	.x86_capability		= { 0 }
65 };
66 
67 EXPORT_SYMBOL(boot_cpu_data);
68 
69 
70 /* Changed in setup_arch, which is called in early boot */
71 static char host_info[(__NEW_UTS_LEN + 1) * 5];
72 
73 static int show_cpuinfo(struct seq_file *m, void *v)
74 {
75 	int i = 0;
76 
77 	seq_printf(m, "processor\t: %d\n", i);
78 	seq_printf(m, "vendor_id\t: User Mode Linux\n");
79 	seq_printf(m, "model name\t: UML\n");
80 	seq_printf(m, "mode\t\t: skas\n");
81 	seq_printf(m, "host\t\t: %s\n", host_info);
82 	seq_printf(m, "fpu\t\t: %s\n", str_yes_no(cpu_has(&boot_cpu_data, X86_FEATURE_FPU)));
83 	seq_printf(m, "flags\t\t:");
84 	for (i = 0; i < 32*NCAPINTS; i++)
85 		if (cpu_has(&boot_cpu_data, i) && (x86_cap_flags[i] != NULL))
86 			seq_printf(m, " %s", x86_cap_flags[i]);
87 	seq_printf(m, "\n");
88 	seq_printf(m, "cache_alignment\t: %d\n", boot_cpu_data.cache_alignment);
89 	seq_printf(m, "bogomips\t: %lu.%02lu\n",
90 		   loops_per_jiffy/(500000/HZ),
91 		   (loops_per_jiffy/(5000/HZ)) % 100);
92 
93 
94 	return 0;
95 }
96 
97 static void *c_start(struct seq_file *m, loff_t *pos)
98 {
99 	return *pos < nr_cpu_ids ? &boot_cpu_data + *pos : NULL;
100 }
101 
102 static void *c_next(struct seq_file *m, void *v, loff_t *pos)
103 {
104 	++*pos;
105 	return c_start(m, pos);
106 }
107 
108 static void c_stop(struct seq_file *m, void *v)
109 {
110 }
111 
112 const struct seq_operations cpuinfo_op = {
113 	.start	= c_start,
114 	.next	= c_next,
115 	.stop	= c_stop,
116 	.show	= show_cpuinfo,
117 };
118 
119 /* Set in linux_main */
120 unsigned long uml_physmem;
121 EXPORT_SYMBOL(uml_physmem);
122 
123 unsigned long uml_reserved; /* Also modified in mem_init */
124 unsigned long start_vm;
125 unsigned long end_vm;
126 
127 /* Set in early boot */
128 static int have_root __initdata;
129 static int have_console __initdata;
130 
131 /* Set in uml_mem_setup and modified in linux_main */
132 unsigned long long physmem_size = 64 * 1024 * 1024;
133 EXPORT_SYMBOL(physmem_size);
134 
135 static const char *usage_string =
136 "User Mode Linux v%s\n"
137 "	available at http://user-mode-linux.sourceforge.net/\n\n";
138 
139 static int __init uml_version_setup(char *line, int *add)
140 {
141 	/* Explicitly use printf() to show version in stdout */
142 	printf("%s\n", init_utsname()->release);
143 	exit(0);
144 
145 	return 0;
146 }
147 
148 __uml_setup("--version", uml_version_setup,
149 "--version\n"
150 "    Prints the version number of the kernel.\n\n"
151 );
152 
153 static int __init uml_root_setup(char *line, int *add)
154 {
155 	have_root = 1;
156 	return 0;
157 }
158 
159 __uml_setup("root=", uml_root_setup,
160 "root=<file containing the root fs>\n"
161 "    This is actually used by the generic kernel in exactly the same\n"
162 "    way as in any other kernel. If you configure a number of block\n"
163 "    devices and want to boot off something other than ubd0, you \n"
164 "    would use something like:\n"
165 "        root=/dev/ubd5\n\n"
166 );
167 
168 static int __init uml_console_setup(char *line, int *add)
169 {
170 	have_console = 1;
171 	return 0;
172 }
173 
174 __uml_setup("console=", uml_console_setup,
175 "console=<preferred console>\n"
176 "    Specify the preferred console output driver\n\n"
177 );
178 
179 static int __init Usage(char *line, int *add)
180 {
181 	const char **p;
182 
183 	printf(usage_string, init_utsname()->release);
184 	p = &__uml_help_start;
185 	/* Explicitly use printf() to show help in stdout */
186 	while (p < &__uml_help_end) {
187 		printf("%s", *p);
188 		p++;
189 	}
190 	exit(0);
191 	return 0;
192 }
193 
194 __uml_setup("--help", Usage,
195 "--help\n"
196 "    Prints this message.\n\n"
197 );
198 
199 static void __init uml_checksetup(char *line, int *add)
200 {
201 	struct uml_param *p;
202 
203 	p = &__uml_setup_start;
204 	while (p < &__uml_setup_end) {
205 		size_t n;
206 
207 		n = strlen(p->str);
208 		if (!strncmp(line, p->str, n) && p->setup_func(line + n, add))
209 			return;
210 		p++;
211 	}
212 }
213 
214 static void __init uml_postsetup(void)
215 {
216 	initcall_t *p;
217 
218 	p = &__uml_postsetup_start;
219 	while (p < &__uml_postsetup_end) {
220 		(*p)();
221 		p++;
222 	}
223 	return;
224 }
225 
226 static int panic_exit(struct notifier_block *self, unsigned long unused1,
227 		      void *unused2)
228 {
229 	kmsg_dump(KMSG_DUMP_PANIC);
230 	bust_spinlocks(1);
231 	bust_spinlocks(0);
232 	uml_exitcode = 1;
233 	os_dump_core();
234 
235 	return NOTIFY_DONE;
236 }
237 
238 static struct notifier_block panic_exit_notifier = {
239 	.notifier_call	= panic_exit,
240 	.priority	= INT_MAX - 1, /* run as 2nd notifier, won't return */
241 };
242 
243 void uml_finishsetup(void)
244 {
245 	cpu_tasks[0] = &init_task;
246 
247 	atomic_notifier_chain_register(&panic_notifier_list,
248 				       &panic_exit_notifier);
249 
250 	uml_postsetup();
251 
252 	new_thread_handler();
253 }
254 
255 /* Set during early boot */
256 unsigned long stub_start;
257 unsigned long task_size;
258 EXPORT_SYMBOL(task_size);
259 
260 unsigned long host_task_size;
261 
262 unsigned long brk_start;
263 unsigned long end_iomem;
264 EXPORT_SYMBOL(end_iomem);
265 
266 #define MIN_VMALLOC (32 * 1024 * 1024)
267 
268 static void __init parse_host_cpu_flags(char *line)
269 {
270 	int i;
271 	for (i = 0; i < 32*NCAPINTS; i++) {
272 		if ((x86_cap_flags[i] != NULL) && strstr(line, x86_cap_flags[i]))
273 			set_cpu_cap(&boot_cpu_data, i);
274 	}
275 }
276 
277 static void __init parse_cache_line(char *line)
278 {
279 	long res;
280 	char *to_parse = strstr(line, ":");
281 	if (to_parse) {
282 		to_parse++;
283 		while (*to_parse != 0 && isspace(*to_parse)) {
284 			to_parse++;
285 		}
286 		if (kstrtoul(to_parse, 10, &res) == 0 && is_power_of_2(res))
287 			boot_cpu_data.cache_alignment = res;
288 		else
289 			boot_cpu_data.cache_alignment = L1_CACHE_BYTES;
290 	}
291 }
292 
293 static unsigned long __init get_top_address(char **envp)
294 {
295 	unsigned long top_addr = (unsigned long) &top_addr;
296 	int i;
297 
298 	/* The earliest variable should be after the program name in ELF */
299 	for (i = 0; envp[i]; i++) {
300 		if ((unsigned long) envp[i] > top_addr)
301 			top_addr = (unsigned long) envp[i];
302 	}
303 
304 	top_addr &= ~(UM_KERN_PAGE_SIZE - 1);
305 	top_addr += UM_KERN_PAGE_SIZE;
306 
307 	return top_addr;
308 }
309 
310 int __init linux_main(int argc, char **argv, char **envp)
311 {
312 	unsigned long avail, diff;
313 	unsigned long virtmem_size, max_physmem;
314 	unsigned long stack;
315 	unsigned int i;
316 	int add;
317 
318 	for (i = 1; i < argc; i++) {
319 		if ((i == 1) && (argv[i][0] == ' '))
320 			continue;
321 		add = 1;
322 		uml_checksetup(argv[i], &add);
323 		if (add)
324 			add_arg(argv[i]);
325 	}
326 	if (have_root == 0)
327 		add_arg(DEFAULT_COMMAND_LINE_ROOT);
328 
329 	if (have_console == 0)
330 		add_arg(DEFAULT_COMMAND_LINE_CONSOLE);
331 
332 	host_task_size = get_top_address(envp);
333 	/* reserve a few pages for the stubs */
334 	stub_start = host_task_size - STUB_DATA_PAGES * PAGE_SIZE;
335 	/* another page for the code portion */
336 	stub_start -= PAGE_SIZE;
337 	host_task_size = stub_start;
338 
339 	/* Limit TASK_SIZE to what is addressable by the page table */
340 	task_size = host_task_size;
341 	if (task_size > (unsigned long long) PTRS_PER_PGD * PGDIR_SIZE)
342 		task_size = PTRS_PER_PGD * PGDIR_SIZE;
343 
344 	/*
345 	 * TASK_SIZE needs to be PGDIR_SIZE aligned or else exit_mmap craps
346 	 * out
347 	 */
348 	task_size = task_size & PGDIR_MASK;
349 
350 	/* OS sanity checks that need to happen before the kernel runs */
351 	os_early_checks();
352 
353 	get_host_cpu_features(parse_host_cpu_flags, parse_cache_line);
354 
355 	brk_start = (unsigned long) sbrk(0);
356 
357 	/*
358 	 * Increase physical memory size for exec-shield users
359 	 * so they actually get what they asked for. This should
360 	 * add zero for non-exec shield users
361 	 */
362 
363 	diff = UML_ROUND_UP(brk_start) - UML_ROUND_UP(&_end);
364 	if (diff > 1024 * 1024) {
365 		os_info("Adding %ld bytes to physical memory to account for "
366 			"exec-shield gap\n", diff);
367 		physmem_size += UML_ROUND_UP(brk_start) - UML_ROUND_UP(&_end);
368 	}
369 
370 	uml_physmem = (unsigned long) __binary_start & PAGE_MASK;
371 
372 	/* Reserve up to 4M after the current brk */
373 	uml_reserved = ROUND_4M(brk_start) + (1 << 22);
374 
375 	setup_machinename(init_utsname()->machine);
376 
377 	physmem_size = (physmem_size + PAGE_SIZE - 1) & PAGE_MASK;
378 	iomem_size = (iomem_size + PAGE_SIZE - 1) & PAGE_MASK;
379 
380 	max_physmem = TASK_SIZE - uml_physmem - iomem_size - MIN_VMALLOC;
381 	if (physmem_size > max_physmem) {
382 		physmem_size = max_physmem;
383 		os_info("Physical memory size shrunk to %llu bytes\n",
384 			physmem_size);
385 	}
386 
387 	high_physmem = uml_physmem + physmem_size;
388 	end_iomem = high_physmem + iomem_size;
389 
390 	start_vm = VMALLOC_START;
391 
392 	virtmem_size = physmem_size;
393 	stack = (unsigned long) argv;
394 	stack &= ~(1024 * 1024 - 1);
395 	avail = stack - start_vm;
396 	if (physmem_size > avail)
397 		virtmem_size = avail;
398 	end_vm = start_vm + virtmem_size;
399 
400 	if (virtmem_size < physmem_size)
401 		os_info("Kernel virtual memory size shrunk to %lu bytes\n",
402 			virtmem_size);
403 
404 	arch_task_struct_size = sizeof(struct task_struct) + host_fp_size;
405 
406 	os_flush_stdout();
407 
408 	return start_uml();
409 }
410 
411 int __init __weak read_initrd(void)
412 {
413 	return 0;
414 }
415 
416 void __init setup_arch(char **cmdline_p)
417 {
418 	u8 rng_seed[32];
419 
420 	stack_protections((unsigned long) init_task.stack);
421 	setup_physmem(uml_physmem, uml_reserved, physmem_size);
422 	uml_dtb_init();
423 	read_initrd();
424 
425 	paging_init();
426 	strscpy(boot_command_line, command_line, COMMAND_LINE_SIZE);
427 	*cmdline_p = command_line;
428 	setup_hostinfo(host_info, sizeof host_info);
429 
430 	if (os_getrandom(rng_seed, sizeof(rng_seed), 0) == sizeof(rng_seed)) {
431 		add_bootloader_randomness(rng_seed, sizeof(rng_seed));
432 		memzero_explicit(rng_seed, sizeof(rng_seed));
433 	}
434 }
435 
436 void __init arch_cpu_finalize_init(void)
437 {
438 	arch_check_bugs();
439 	os_check_bugs();
440 }
441 
442 void apply_seal_endbr(s32 *start, s32 *end)
443 {
444 }
445 
446 void apply_retpolines(s32 *start, s32 *end)
447 {
448 }
449 
450 void apply_returns(s32 *start, s32 *end)
451 {
452 }
453 
454 void apply_fineibt(s32 *start_retpoline, s32 *end_retpoline,
455 		   s32 *start_cfi, s32 *end_cfi)
456 {
457 }
458 
459 void apply_alternatives(struct alt_instr *start, struct alt_instr *end)
460 {
461 }
462 
463 void *text_poke(void *addr, const void *opcode, size_t len)
464 {
465 	/*
466 	 * In UML, the only reference to this function is in
467 	 * apply_relocate_add(), which shouldn't ever actually call this
468 	 * because UML doesn't have live patching.
469 	 */
470 	WARN_ON(1);
471 
472 	return memcpy(addr, opcode, len);
473 }
474 
475 void *text_poke_copy(void *addr, const void *opcode, size_t len)
476 {
477 	return text_poke(addr, opcode, len);
478 }
479 
480 void smp_text_poke_sync_each_cpu(void)
481 {
482 }
483 
484 void uml_pm_wake(void)
485 {
486 	pm_system_wakeup();
487 }
488 
489 #ifdef CONFIG_PM_SLEEP
490 static int um_suspend_valid(suspend_state_t state)
491 {
492 	return state == PM_SUSPEND_MEM;
493 }
494 
495 static int um_suspend_prepare(void)
496 {
497 	um_irqs_suspend();
498 	return 0;
499 }
500 
501 static int um_suspend_enter(suspend_state_t state)
502 {
503 	if (WARN_ON(state != PM_SUSPEND_MEM))
504 		return -EINVAL;
505 
506 	/*
507 	 * This is identical to the idle sleep, but we've just
508 	 * (during suspend) turned off all interrupt sources
509 	 * except for the ones we want, so now we can only wake
510 	 * up on something we actually want to wake up on. All
511 	 * timing has also been suspended.
512 	 */
513 	um_idle_sleep();
514 	return 0;
515 }
516 
517 static void um_suspend_finish(void)
518 {
519 	um_irqs_resume();
520 }
521 
522 const struct platform_suspend_ops um_suspend_ops = {
523 	.valid = um_suspend_valid,
524 	.prepare = um_suspend_prepare,
525 	.enter = um_suspend_enter,
526 	.finish = um_suspend_finish,
527 };
528 
529 static int init_pm_wake_signal(void)
530 {
531 	/*
532 	 * In external time-travel mode we can't use signals to wake up
533 	 * since that would mess with the scheduling. We'll have to do
534 	 * some additional work to support wakeup on virtio devices or
535 	 * similar, perhaps implementing a fake RTC controller that can
536 	 * trigger wakeup (and request the appropriate scheduling from
537 	 * the external scheduler when going to suspend.)
538 	 */
539 	if (time_travel_mode != TT_MODE_EXTERNAL)
540 		register_pm_wake_signal();
541 
542 	suspend_set_ops(&um_suspend_ops);
543 
544 	return 0;
545 }
546 
547 late_initcall(init_pm_wake_signal);
548 #endif
549