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/utsname.h>
16 #include <linux/sched.h>
17 #include <linux/sched/task.h>
18 #include <linux/kmsg_dump.h>
19 #include <linux/suspend.h>
20 #include <linux/random.h>
21
22 #include <asm/processor.h>
23 #include <asm/cpufeature.h>
24 #include <asm/sections.h>
25 #include <asm/setup.h>
26 #include <as-layout.h>
27 #include <arch.h>
28 #include <init.h>
29 #include <kern.h>
30 #include <kern_util.h>
31 #include <mem_user.h>
32 #include <os.h>
33
34 #define DEFAULT_COMMAND_LINE_ROOT "root=98:0"
35 #define DEFAULT_COMMAND_LINE_CONSOLE "console=tty0"
36
37 /* Changed in add_arg and setup_arch, which run before SMP is started */
38 static char __initdata command_line[COMMAND_LINE_SIZE] = { 0 };
39
add_arg(char * arg)40 static void __init add_arg(char *arg)
41 {
42 if (strlen(command_line) + strlen(arg) + 1 > COMMAND_LINE_SIZE) {
43 os_warn("add_arg: Too many command line arguments!\n");
44 exit(1);
45 }
46 if (strlen(command_line) > 0)
47 strcat(command_line, " ");
48 strcat(command_line, arg);
49 }
50
51 /*
52 * These fields are initialized at boot time and not changed.
53 * XXX This structure is used only in the non-SMP case. Maybe this
54 * should be moved to smp.c.
55 */
56 struct cpuinfo_um boot_cpu_data = {
57 .loops_per_jiffy = 0,
58 .ipi_pipe = { -1, -1 },
59 .cache_alignment = L1_CACHE_BYTES,
60 .x86_capability = { 0 }
61 };
62
63 EXPORT_SYMBOL(boot_cpu_data);
64
65 union thread_union cpu0_irqstack
66 __section(".data..init_irqstack") =
67 { .thread_info = INIT_THREAD_INFO(init_task) };
68
69 /* Changed in setup_arch, which is called in early boot */
70 static char host_info[(__NEW_UTS_LEN + 1) * 5];
71
show_cpuinfo(struct seq_file * m,void * v)72 static int show_cpuinfo(struct seq_file *m, void *v)
73 {
74 int i = 0;
75
76 seq_printf(m, "processor\t: %d\n", i);
77 seq_printf(m, "vendor_id\t: User Mode Linux\n");
78 seq_printf(m, "model name\t: UML\n");
79 seq_printf(m, "mode\t\t: skas\n");
80 seq_printf(m, "host\t\t: %s\n", host_info);
81 seq_printf(m, "fpu\t\t: %s\n", cpu_has(&boot_cpu_data, X86_FEATURE_FPU) ? "yes" : "no");
82 seq_printf(m, "flags\t\t:");
83 for (i = 0; i < 32*NCAPINTS; i++)
84 if (cpu_has(&boot_cpu_data, i) && (x86_cap_flags[i] != NULL))
85 seq_printf(m, " %s", x86_cap_flags[i]);
86 seq_printf(m, "\n");
87 seq_printf(m, "cache_alignment\t: %d\n", boot_cpu_data.cache_alignment);
88 seq_printf(m, "bogomips\t: %lu.%02lu\n",
89 loops_per_jiffy/(500000/HZ),
90 (loops_per_jiffy/(5000/HZ)) % 100);
91
92
93 return 0;
94 }
95
c_start(struct seq_file * m,loff_t * pos)96 static void *c_start(struct seq_file *m, loff_t *pos)
97 {
98 return *pos < nr_cpu_ids ? cpu_data + *pos : NULL;
99 }
100
c_next(struct seq_file * m,void * v,loff_t * pos)101 static void *c_next(struct seq_file *m, void *v, loff_t *pos)
102 {
103 ++*pos;
104 return c_start(m, pos);
105 }
106
c_stop(struct seq_file * m,void * v)107 static void c_stop(struct seq_file *m, void *v)
108 {
109 }
110
111 const struct seq_operations cpuinfo_op = {
112 .start = c_start,
113 .next = c_next,
114 .stop = c_stop,
115 .show = show_cpuinfo,
116 };
117
118 /* Set in linux_main */
119 unsigned long uml_physmem;
120 EXPORT_SYMBOL(uml_physmem);
121
122 unsigned long uml_reserved; /* Also modified in mem_init */
123 unsigned long start_vm;
124 unsigned long end_vm;
125
126 /* Set in uml_ncpus_setup */
127 int ncpus = 1;
128
129 /* Set in early boot */
130 static int have_root __initdata;
131 static int have_console __initdata;
132
133 /* Set in uml_mem_setup and modified in linux_main */
134 long long physmem_size = 32 * 1024 * 1024;
135 EXPORT_SYMBOL(physmem_size);
136
137 static const char *usage_string =
138 "User Mode Linux v%s\n"
139 " available at http://user-mode-linux.sourceforge.net/\n\n";
140
uml_version_setup(char * line,int * add)141 static int __init uml_version_setup(char *line, int *add)
142 {
143 /* Explicitly use printf() to show version in stdout */
144 printf("%s\n", init_utsname()->release);
145 exit(0);
146
147 return 0;
148 }
149
150 __uml_setup("--version", uml_version_setup,
151 "--version\n"
152 " Prints the version number of the kernel.\n\n"
153 );
154
uml_root_setup(char * line,int * add)155 static int __init uml_root_setup(char *line, int *add)
156 {
157 have_root = 1;
158 return 0;
159 }
160
161 __uml_setup("root=", uml_root_setup,
162 "root=<file containing the root fs>\n"
163 " This is actually used by the generic kernel in exactly the same\n"
164 " way as in any other kernel. If you configure a number of block\n"
165 " devices and want to boot off something other than ubd0, you \n"
166 " would use something like:\n"
167 " root=/dev/ubd5\n\n"
168 );
169
no_skas_debug_setup(char * line,int * add)170 static int __init no_skas_debug_setup(char *line, int *add)
171 {
172 os_warn("'debug' is not necessary to gdb UML in skas mode - run\n");
173 os_warn("'gdb linux'\n");
174
175 return 0;
176 }
177
178 __uml_setup("debug", no_skas_debug_setup,
179 "debug\n"
180 " this flag is not needed to run gdb on UML in skas mode\n\n"
181 );
182
uml_console_setup(char * line,int * add)183 static int __init uml_console_setup(char *line, int *add)
184 {
185 have_console = 1;
186 return 0;
187 }
188
189 __uml_setup("console=", uml_console_setup,
190 "console=<preferred console>\n"
191 " Specify the preferred console output driver\n\n"
192 );
193
Usage(char * line,int * add)194 static int __init Usage(char *line, int *add)
195 {
196 const char **p;
197
198 printf(usage_string, init_utsname()->release);
199 p = &__uml_help_start;
200 /* Explicitly use printf() to show help in stdout */
201 while (p < &__uml_help_end) {
202 printf("%s", *p);
203 p++;
204 }
205 exit(0);
206 return 0;
207 }
208
209 __uml_setup("--help", Usage,
210 "--help\n"
211 " Prints this message.\n\n"
212 );
213
uml_checksetup(char * line,int * add)214 static void __init uml_checksetup(char *line, int *add)
215 {
216 struct uml_param *p;
217
218 p = &__uml_setup_start;
219 while (p < &__uml_setup_end) {
220 size_t n;
221
222 n = strlen(p->str);
223 if (!strncmp(line, p->str, n) && p->setup_func(line + n, add))
224 return;
225 p++;
226 }
227 }
228
uml_postsetup(void)229 static void __init uml_postsetup(void)
230 {
231 initcall_t *p;
232
233 p = &__uml_postsetup_start;
234 while (p < &__uml_postsetup_end) {
235 (*p)();
236 p++;
237 }
238 return;
239 }
240
panic_exit(struct notifier_block * self,unsigned long unused1,void * unused2)241 static int panic_exit(struct notifier_block *self, unsigned long unused1,
242 void *unused2)
243 {
244 kmsg_dump(KMSG_DUMP_PANIC);
245 bust_spinlocks(1);
246 bust_spinlocks(0);
247 uml_exitcode = 1;
248 os_dump_core();
249 return 0;
250 }
251
252 static struct notifier_block panic_exit_notifier = {
253 .notifier_call = panic_exit,
254 .next = NULL,
255 .priority = 0
256 };
257
uml_finishsetup(void)258 void uml_finishsetup(void)
259 {
260 atomic_notifier_chain_register(&panic_notifier_list,
261 &panic_exit_notifier);
262
263 uml_postsetup();
264
265 new_thread_handler();
266 }
267
268 /* Set during early boot */
269 unsigned long stub_start;
270 unsigned long task_size;
271 EXPORT_SYMBOL(task_size);
272
273 unsigned long host_task_size;
274
275 unsigned long brk_start;
276 unsigned long end_iomem;
277 EXPORT_SYMBOL(end_iomem);
278
279 #define MIN_VMALLOC (32 * 1024 * 1024)
280
parse_host_cpu_flags(char * line)281 static void parse_host_cpu_flags(char *line)
282 {
283 int i;
284 for (i = 0; i < 32*NCAPINTS; i++) {
285 if ((x86_cap_flags[i] != NULL) && strstr(line, x86_cap_flags[i]))
286 set_cpu_cap(&boot_cpu_data, i);
287 }
288 }
parse_cache_line(char * line)289 static void parse_cache_line(char *line)
290 {
291 long res;
292 char *to_parse = strstr(line, ":");
293 if (to_parse) {
294 to_parse++;
295 while (*to_parse != 0 && isspace(*to_parse)) {
296 to_parse++;
297 }
298 if (kstrtoul(to_parse, 10, &res) == 0 && is_power_of_2(res))
299 boot_cpu_data.cache_alignment = res;
300 else
301 boot_cpu_data.cache_alignment = L1_CACHE_BYTES;
302 }
303 }
304
linux_main(int argc,char ** argv)305 int __init linux_main(int argc, char **argv)
306 {
307 unsigned long avail, diff;
308 unsigned long virtmem_size, max_physmem;
309 unsigned long stack;
310 unsigned int i;
311 int add;
312
313 for (i = 1; i < argc; i++) {
314 if ((i == 1) && (argv[i][0] == ' '))
315 continue;
316 add = 1;
317 uml_checksetup(argv[i], &add);
318 if (add)
319 add_arg(argv[i]);
320 }
321 if (have_root == 0)
322 add_arg(DEFAULT_COMMAND_LINE_ROOT);
323
324 if (have_console == 0)
325 add_arg(DEFAULT_COMMAND_LINE_CONSOLE);
326
327 host_task_size = os_get_top_address();
328 /* reserve two pages for the stubs */
329 host_task_size -= 2 * PAGE_SIZE;
330 stub_start = host_task_size;
331
332 /*
333 * TASK_SIZE needs to be PGDIR_SIZE aligned or else exit_mmap craps
334 * out
335 */
336 task_size = host_task_size & PGDIR_MASK;
337
338 /* OS sanity checks that need to happen before the kernel runs */
339 os_early_checks();
340
341 get_host_cpu_features(parse_host_cpu_flags, parse_cache_line);
342
343 brk_start = (unsigned long) sbrk(0);
344
345 /*
346 * Increase physical memory size for exec-shield users
347 * so they actually get what they asked for. This should
348 * add zero for non-exec shield users
349 */
350
351 diff = UML_ROUND_UP(brk_start) - UML_ROUND_UP(&_end);
352 if (diff > 1024 * 1024) {
353 os_info("Adding %ld bytes to physical memory to account for "
354 "exec-shield gap\n", diff);
355 physmem_size += UML_ROUND_UP(brk_start) - UML_ROUND_UP(&_end);
356 }
357
358 uml_physmem = (unsigned long) __binary_start & PAGE_MASK;
359
360 /* Reserve up to 4M after the current brk */
361 uml_reserved = ROUND_4M(brk_start) + (1 << 22);
362
363 setup_machinename(init_utsname()->machine);
364
365 highmem = 0;
366 iomem_size = (iomem_size + PAGE_SIZE - 1) & PAGE_MASK;
367 max_physmem = TASK_SIZE - uml_physmem - iomem_size - MIN_VMALLOC;
368
369 /*
370 * Zones have to begin on a 1 << MAX_ORDER page boundary,
371 * so this makes sure that's true for highmem
372 */
373 max_physmem &= ~((1 << (PAGE_SHIFT + MAX_ORDER)) - 1);
374 if (physmem_size + iomem_size > max_physmem) {
375 highmem = physmem_size + iomem_size - max_physmem;
376 physmem_size -= highmem;
377 }
378
379 high_physmem = uml_physmem + physmem_size;
380 end_iomem = high_physmem + iomem_size;
381 high_memory = (void *) end_iomem;
382
383 start_vm = VMALLOC_START;
384
385 virtmem_size = physmem_size;
386 stack = (unsigned long) argv;
387 stack &= ~(1024 * 1024 - 1);
388 avail = stack - start_vm;
389 if (physmem_size > avail)
390 virtmem_size = avail;
391 end_vm = start_vm + virtmem_size;
392
393 if (virtmem_size < physmem_size)
394 os_info("Kernel virtual memory size shrunk to %lu bytes\n",
395 virtmem_size);
396
397 os_flush_stdout();
398
399 return start_uml();
400 }
401
read_initrd(void)402 int __init __weak read_initrd(void)
403 {
404 return 0;
405 }
406
setup_arch(char ** cmdline_p)407 void __init setup_arch(char **cmdline_p)
408 {
409 u8 rng_seed[32];
410
411 stack_protections((unsigned long) &init_thread_info);
412 setup_physmem(uml_physmem, uml_reserved, physmem_size, highmem);
413 mem_total_pages(physmem_size, iomem_size, highmem);
414 read_initrd();
415
416 paging_init();
417 strlcpy(boot_command_line, command_line, COMMAND_LINE_SIZE);
418 *cmdline_p = command_line;
419 setup_hostinfo(host_info, sizeof host_info);
420
421 if (os_getrandom(rng_seed, sizeof(rng_seed), 0) == sizeof(rng_seed)) {
422 add_bootloader_randomness(rng_seed, sizeof(rng_seed));
423 memzero_explicit(rng_seed, sizeof(rng_seed));
424 }
425 }
426
arch_cpu_finalize_init(void)427 void __init arch_cpu_finalize_init(void)
428 {
429 arch_check_bugs();
430 os_check_bugs();
431 }
432
apply_retpolines(s32 * start,s32 * end)433 void apply_retpolines(s32 *start, s32 *end)
434 {
435 }
436
apply_returns(s32 * start,s32 * end)437 void apply_returns(s32 *start, s32 *end)
438 {
439 }
440
apply_alternatives(struct alt_instr * start,struct alt_instr * end)441 void apply_alternatives(struct alt_instr *start, struct alt_instr *end)
442 {
443 }
444
text_poke(void * addr,const void * opcode,size_t len)445 void *text_poke(void *addr, const void *opcode, size_t len)
446 {
447 /*
448 * In UML, the only reference to this function is in
449 * apply_relocate_add(), which shouldn't ever actually call this
450 * because UML doesn't have live patching.
451 */
452 WARN_ON(1);
453
454 return memcpy(addr, opcode, len);
455 }
456
text_poke_sync(void)457 void text_poke_sync(void)
458 {
459 }
460
uml_pm_wake(void)461 void uml_pm_wake(void)
462 {
463 pm_system_wakeup();
464 }
465
466 #ifdef CONFIG_PM_SLEEP
um_suspend_valid(suspend_state_t state)467 static int um_suspend_valid(suspend_state_t state)
468 {
469 return state == PM_SUSPEND_MEM;
470 }
471
um_suspend_prepare(void)472 static int um_suspend_prepare(void)
473 {
474 um_irqs_suspend();
475 return 0;
476 }
477
um_suspend_enter(suspend_state_t state)478 static int um_suspend_enter(suspend_state_t state)
479 {
480 if (WARN_ON(state != PM_SUSPEND_MEM))
481 return -EINVAL;
482
483 /*
484 * This is identical to the idle sleep, but we've just
485 * (during suspend) turned off all interrupt sources
486 * except for the ones we want, so now we can only wake
487 * up on something we actually want to wake up on. All
488 * timing has also been suspended.
489 */
490 um_idle_sleep();
491 return 0;
492 }
493
um_suspend_finish(void)494 static void um_suspend_finish(void)
495 {
496 um_irqs_resume();
497 }
498
499 const struct platform_suspend_ops um_suspend_ops = {
500 .valid = um_suspend_valid,
501 .prepare = um_suspend_prepare,
502 .enter = um_suspend_enter,
503 .finish = um_suspend_finish,
504 };
505
init_pm_wake_signal(void)506 static int init_pm_wake_signal(void)
507 {
508 /*
509 * In external time-travel mode we can't use signals to wake up
510 * since that would mess with the scheduling. We'll have to do
511 * some additional work to support wakeup on virtio devices or
512 * similar, perhaps implementing a fake RTC controller that can
513 * trigger wakeup (and request the appropriate scheduling from
514 * the external scheduler when going to suspend.)
515 */
516 if (time_travel_mode != TT_MODE_EXTERNAL)
517 register_pm_wake_signal();
518
519 suspend_set_ops(&um_suspend_ops);
520
521 return 0;
522 }
523
524 late_initcall(init_pm_wake_signal);
525 #endif
526