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/delay.h>
7 #include <linux/init.h>
8 #include <linux/mm.h>
9 #include <linux/ctype.h>
10 #include <linux/module.h>
11 #include <linux/panic_notifier.h>
12 #include <linux/seq_file.h>
13 #include <linux/string.h>
14 #include <linux/utsname.h>
15 #include <linux/sched.h>
16 #include <linux/sched/task.h>
17 #include <linux/kmsg_dump.h>
18 #include <linux/suspend.h>
19 #include <linux/random.h>
20
21 #include <asm/processor.h>
22 #include <asm/cpufeature.h>
23 #include <asm/sections.h>
24 #include <asm/setup.h>
25 #include <as-layout.h>
26 #include <arch.h>
27 #include <init.h>
28 #include <kern.h>
29 #include <kern_util.h>
30 #include <mem_user.h>
31 #include <os.h>
32
33 #include "um_arch.h"
34
35 #define DEFAULT_COMMAND_LINE_ROOT "root=98:0"
36 #define DEFAULT_COMMAND_LINE_CONSOLE "console=tty0"
37
38 /* Changed in add_arg and setup_arch, which run before SMP is started */
39 static char __initdata command_line[COMMAND_LINE_SIZE] = { 0 };
40
add_arg(char * arg)41 static void __init add_arg(char *arg)
42 {
43 if (strlen(command_line) + strlen(arg) + 1 > COMMAND_LINE_SIZE) {
44 os_warn("add_arg: Too many command line arguments!\n");
45 exit(1);
46 }
47 if (strlen(command_line) > 0)
48 strcat(command_line, " ");
49 strcat(command_line, arg);
50 }
51
52 /*
53 * These fields are initialized at boot time and not changed.
54 * XXX This structure is used only in the non-SMP case. Maybe this
55 * should be moved to smp.c.
56 */
57 struct cpuinfo_um boot_cpu_data = {
58 .loops_per_jiffy = 0,
59 .ipi_pipe = { -1, -1 },
60 .cache_alignment = L1_CACHE_BYTES,
61 .x86_capability = { 0 }
62 };
63
64 EXPORT_SYMBOL(boot_cpu_data);
65
66 union thread_union cpu0_irqstack
67 __section(".data..init_irqstack") =
68 { .thread_info = INIT_THREAD_INFO(init_task) };
69
70 /* Changed in setup_arch, which is called in early boot */
71 static char host_info[(__NEW_UTS_LEN + 1) * 5];
72
show_cpuinfo(struct seq_file * m,void * v)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", cpu_has(&boot_cpu_data, X86_FEATURE_FPU) ? "yes" : "no");
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
c_start(struct seq_file * m,loff_t * pos)97 static void *c_start(struct seq_file *m, loff_t *pos)
98 {
99 return *pos < nr_cpu_ids ? cpu_data + *pos : NULL;
100 }
101
c_next(struct seq_file * m,void * v,loff_t * pos)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
c_stop(struct seq_file * m,void * v)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 uml_ncpus_setup */
128 int ncpus = 1;
129
130 /* Set in early boot */
131 static int have_root __initdata;
132 static int have_console __initdata;
133
134 /* Set in uml_mem_setup and modified in linux_main */
135 long long physmem_size = 64 * 1024 * 1024;
136 EXPORT_SYMBOL(physmem_size);
137
138 static const char *usage_string =
139 "User Mode Linux v%s\n"
140 " available at http://user-mode-linux.sourceforge.net/\n\n";
141
uml_version_setup(char * line,int * add)142 static int __init uml_version_setup(char *line, int *add)
143 {
144 /* Explicitly use printf() to show version in stdout */
145 printf("%s\n", init_utsname()->release);
146 exit(0);
147
148 return 0;
149 }
150
151 __uml_setup("--version", uml_version_setup,
152 "--version\n"
153 " Prints the version number of the kernel.\n\n"
154 );
155
uml_root_setup(char * line,int * add)156 static int __init uml_root_setup(char *line, int *add)
157 {
158 have_root = 1;
159 return 0;
160 }
161
162 __uml_setup("root=", uml_root_setup,
163 "root=<file containing the root fs>\n"
164 " This is actually used by the generic kernel in exactly the same\n"
165 " way as in any other kernel. If you configure a number of block\n"
166 " devices and want to boot off something other than ubd0, you \n"
167 " would use something like:\n"
168 " root=/dev/ubd5\n\n"
169 );
170
no_skas_debug_setup(char * line,int * add)171 static int __init no_skas_debug_setup(char *line, int *add)
172 {
173 os_warn("'debug' is not necessary to gdb UML in skas mode - run\n");
174 os_warn("'gdb linux'\n");
175
176 return 0;
177 }
178
179 __uml_setup("debug", no_skas_debug_setup,
180 "debug\n"
181 " this flag is not needed to run gdb on UML in skas mode\n\n"
182 );
183
uml_console_setup(char * line,int * add)184 static int __init uml_console_setup(char *line, int *add)
185 {
186 have_console = 1;
187 return 0;
188 }
189
190 __uml_setup("console=", uml_console_setup,
191 "console=<preferred console>\n"
192 " Specify the preferred console output driver\n\n"
193 );
194
Usage(char * line,int * add)195 static int __init Usage(char *line, int *add)
196 {
197 const char **p;
198
199 printf(usage_string, init_utsname()->release);
200 p = &__uml_help_start;
201 /* Explicitly use printf() to show help in stdout */
202 while (p < &__uml_help_end) {
203 printf("%s", *p);
204 p++;
205 }
206 exit(0);
207 return 0;
208 }
209
210 __uml_setup("--help", Usage,
211 "--help\n"
212 " Prints this message.\n\n"
213 );
214
uml_checksetup(char * line,int * add)215 static void __init uml_checksetup(char *line, int *add)
216 {
217 struct uml_param *p;
218
219 p = &__uml_setup_start;
220 while (p < &__uml_setup_end) {
221 size_t n;
222
223 n = strlen(p->str);
224 if (!strncmp(line, p->str, n) && p->setup_func(line + n, add))
225 return;
226 p++;
227 }
228 }
229
uml_postsetup(void)230 static void __init uml_postsetup(void)
231 {
232 initcall_t *p;
233
234 p = &__uml_postsetup_start;
235 while (p < &__uml_postsetup_end) {
236 (*p)();
237 p++;
238 }
239 return;
240 }
241
panic_exit(struct notifier_block * self,unsigned long unused1,void * unused2)242 static int panic_exit(struct notifier_block *self, unsigned long unused1,
243 void *unused2)
244 {
245 kmsg_dump(KMSG_DUMP_PANIC);
246 bust_spinlocks(1);
247 bust_spinlocks(0);
248 uml_exitcode = 1;
249 os_dump_core();
250
251 return NOTIFY_DONE;
252 }
253
254 static struct notifier_block panic_exit_notifier = {
255 .notifier_call = panic_exit,
256 .priority = INT_MAX - 1, /* run as 2nd notifier, won't return */
257 };
258
uml_finishsetup(void)259 void uml_finishsetup(void)
260 {
261 atomic_notifier_chain_register(&panic_notifier_list,
262 &panic_exit_notifier);
263
264 uml_postsetup();
265
266 new_thread_handler();
267 }
268
269 /* Set during early boot */
270 unsigned long stub_start;
271 unsigned long task_size;
272 EXPORT_SYMBOL(task_size);
273
274 unsigned long host_task_size;
275
276 unsigned long brk_start;
277 unsigned long end_iomem;
278 EXPORT_SYMBOL(end_iomem);
279
280 #define MIN_VMALLOC (32 * 1024 * 1024)
281
parse_host_cpu_flags(char * line)282 static void parse_host_cpu_flags(char *line)
283 {
284 int i;
285 for (i = 0; i < 32*NCAPINTS; i++) {
286 if ((x86_cap_flags[i] != NULL) && strstr(line, x86_cap_flags[i]))
287 set_cpu_cap(&boot_cpu_data, i);
288 }
289 }
parse_cache_line(char * line)290 static void parse_cache_line(char *line)
291 {
292 long res;
293 char *to_parse = strstr(line, ":");
294 if (to_parse) {
295 to_parse++;
296 while (*to_parse != 0 && isspace(*to_parse)) {
297 to_parse++;
298 }
299 if (kstrtoul(to_parse, 10, &res) == 0 && is_power_of_2(res))
300 boot_cpu_data.cache_alignment = res;
301 else
302 boot_cpu_data.cache_alignment = L1_CACHE_BYTES;
303 }
304 }
305
linux_main(int argc,char ** argv)306 int __init linux_main(int argc, char **argv)
307 {
308 unsigned long avail, diff;
309 unsigned long virtmem_size, max_physmem;
310 unsigned long stack;
311 unsigned int i;
312 int add;
313
314 for (i = 1; i < argc; i++) {
315 if ((i == 1) && (argv[i][0] == ' '))
316 continue;
317 add = 1;
318 uml_checksetup(argv[i], &add);
319 if (add)
320 add_arg(argv[i]);
321 }
322 if (have_root == 0)
323 add_arg(DEFAULT_COMMAND_LINE_ROOT);
324
325 if (have_console == 0)
326 add_arg(DEFAULT_COMMAND_LINE_CONSOLE);
327
328 host_task_size = os_get_top_address();
329 /* reserve two pages for the stubs */
330 host_task_size -= 2 * PAGE_SIZE;
331 stub_start = host_task_size;
332
333 /*
334 * TASK_SIZE needs to be PGDIR_SIZE aligned or else exit_mmap craps
335 * out
336 */
337 task_size = host_task_size & PGDIR_MASK;
338
339 /* OS sanity checks that need to happen before the kernel runs */
340 os_early_checks();
341
342 get_host_cpu_features(parse_host_cpu_flags, parse_cache_line);
343
344 brk_start = (unsigned long) sbrk(0);
345
346 /*
347 * Increase physical memory size for exec-shield users
348 * so they actually get what they asked for. This should
349 * add zero for non-exec shield users
350 */
351
352 diff = UML_ROUND_UP(brk_start) - UML_ROUND_UP(&_end);
353 if (diff > 1024 * 1024) {
354 os_info("Adding %ld bytes to physical memory to account for "
355 "exec-shield gap\n", diff);
356 physmem_size += UML_ROUND_UP(brk_start) - UML_ROUND_UP(&_end);
357 }
358
359 uml_physmem = (unsigned long) __binary_start & PAGE_MASK;
360
361 /* Reserve up to 4M after the current brk */
362 uml_reserved = ROUND_4M(brk_start) + (1 << 22);
363
364 setup_machinename(init_utsname()->machine);
365
366 highmem = 0;
367 iomem_size = (iomem_size + PAGE_SIZE - 1) & PAGE_MASK;
368 max_physmem = TASK_SIZE - uml_physmem - iomem_size - MIN_VMALLOC;
369
370 /*
371 * Zones have to begin on a 1 << MAX_ORDER page boundary,
372 * so this makes sure that's true for highmem
373 */
374 max_physmem &= ~((1 << (PAGE_SHIFT + MAX_ORDER)) - 1);
375 if (physmem_size + iomem_size > max_physmem) {
376 highmem = physmem_size + iomem_size - max_physmem;
377 physmem_size -= highmem;
378 }
379
380 high_physmem = uml_physmem + physmem_size;
381 end_iomem = high_physmem + iomem_size;
382 high_memory = (void *) end_iomem;
383
384 start_vm = VMALLOC_START;
385
386 virtmem_size = physmem_size;
387 stack = (unsigned long) argv;
388 stack &= ~(1024 * 1024 - 1);
389 avail = stack - start_vm;
390 if (physmem_size > avail)
391 virtmem_size = avail;
392 end_vm = start_vm + virtmem_size;
393
394 if (virtmem_size < physmem_size)
395 os_info("Kernel virtual memory size shrunk to %lu bytes\n",
396 virtmem_size);
397
398 os_flush_stdout();
399
400 return start_uml();
401 }
402
read_initrd(void)403 int __init __weak read_initrd(void)
404 {
405 return 0;
406 }
407
setup_arch(char ** cmdline_p)408 void __init setup_arch(char **cmdline_p)
409 {
410 u8 rng_seed[32];
411
412 stack_protections((unsigned long) &init_thread_info);
413 setup_physmem(uml_physmem, uml_reserved, physmem_size, highmem);
414 mem_total_pages(physmem_size, iomem_size, highmem);
415 uml_dtb_init();
416 read_initrd();
417
418 paging_init();
419 strscpy(boot_command_line, command_line, COMMAND_LINE_SIZE);
420 *cmdline_p = command_line;
421 setup_hostinfo(host_info, sizeof host_info);
422
423 if (os_getrandom(rng_seed, sizeof(rng_seed), 0) == sizeof(rng_seed)) {
424 add_bootloader_randomness(rng_seed, sizeof(rng_seed));
425 memzero_explicit(rng_seed, sizeof(rng_seed));
426 }
427 }
428
check_bugs(void)429 void __init check_bugs(void)
430 {
431 arch_check_bugs();
432 os_check_bugs();
433 }
434
apply_ibt_endbr(s32 * start,s32 * end)435 void apply_ibt_endbr(s32 *start, s32 *end)
436 {
437 }
438
apply_retpolines(s32 * start,s32 * end)439 void apply_retpolines(s32 *start, s32 *end)
440 {
441 }
442
apply_returns(s32 * start,s32 * end)443 void apply_returns(s32 *start, s32 *end)
444 {
445 }
446
apply_alternatives(struct alt_instr * start,struct alt_instr * end)447 void apply_alternatives(struct alt_instr *start, struct alt_instr *end)
448 {
449 }
450
text_poke(void * addr,const void * opcode,size_t len)451 void *text_poke(void *addr, const void *opcode, size_t len)
452 {
453 /*
454 * In UML, the only reference to this function is in
455 * apply_relocate_add(), which shouldn't ever actually call this
456 * because UML doesn't have live patching.
457 */
458 WARN_ON(1);
459
460 return memcpy(addr, opcode, len);
461 }
462
text_poke_sync(void)463 void text_poke_sync(void)
464 {
465 }
466
uml_pm_wake(void)467 void uml_pm_wake(void)
468 {
469 pm_system_wakeup();
470 }
471
472 #ifdef CONFIG_PM_SLEEP
um_suspend_valid(suspend_state_t state)473 static int um_suspend_valid(suspend_state_t state)
474 {
475 return state == PM_SUSPEND_MEM;
476 }
477
um_suspend_prepare(void)478 static int um_suspend_prepare(void)
479 {
480 um_irqs_suspend();
481 return 0;
482 }
483
um_suspend_enter(suspend_state_t state)484 static int um_suspend_enter(suspend_state_t state)
485 {
486 if (WARN_ON(state != PM_SUSPEND_MEM))
487 return -EINVAL;
488
489 /*
490 * This is identical to the idle sleep, but we've just
491 * (during suspend) turned off all interrupt sources
492 * except for the ones we want, so now we can only wake
493 * up on something we actually want to wake up on. All
494 * timing has also been suspended.
495 */
496 um_idle_sleep();
497 return 0;
498 }
499
um_suspend_finish(void)500 static void um_suspend_finish(void)
501 {
502 um_irqs_resume();
503 }
504
505 const struct platform_suspend_ops um_suspend_ops = {
506 .valid = um_suspend_valid,
507 .prepare = um_suspend_prepare,
508 .enter = um_suspend_enter,
509 .finish = um_suspend_finish,
510 };
511
init_pm_wake_signal(void)512 static int init_pm_wake_signal(void)
513 {
514 /*
515 * In external time-travel mode we can't use signals to wake up
516 * since that would mess with the scheduling. We'll have to do
517 * some additional work to support wakeup on virtio devices or
518 * similar, perhaps implementing a fake RTC controller that can
519 * trigger wakeup (and request the appropriate scheduling from
520 * the external scheduler when going to suspend.)
521 */
522 if (time_travel_mode != TT_MODE_EXTERNAL)
523 register_pm_wake_signal();
524
525 suspend_set_ops(&um_suspend_ops);
526
527 return 0;
528 }
529
530 late_initcall(init_pm_wake_signal);
531 #endif
532