Commit | Line | Data |
---|---|---|
995473ae | 1 | /* |
ba180fd4 | 2 | * Copyright (C) 2000 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com) |
1da177e4 LT |
3 | * Copyright 2003 PathScale, Inc. |
4 | * Licensed under the GPL | |
5 | */ | |
6 | ||
c5d4bb17 JD |
7 | #include <linux/stddef.h> |
8 | #include <linux/err.h> | |
9 | #include <linux/hardirq.h> | |
c5d4bb17 | 10 | #include <linux/mm.h> |
6613c5e8 | 11 | #include <linux/module.h> |
c5d4bb17 JD |
12 | #include <linux/personality.h> |
13 | #include <linux/proc_fs.h> | |
14 | #include <linux/ptrace.h> | |
15 | #include <linux/random.h> | |
5a0e3ad6 | 16 | #include <linux/slab.h> |
c5d4bb17 | 17 | #include <linux/sched.h> |
6613c5e8 | 18 | #include <linux/seq_file.h> |
c5d4bb17 JD |
19 | #include <linux/tick.h> |
20 | #include <linux/threads.h> | |
d50349b0 | 21 | #include <linux/tracehook.h> |
c5d4bb17 JD |
22 | #include <asm/current.h> |
23 | #include <asm/pgtable.h> | |
445c5786 | 24 | #include <asm/mmu_context.h> |
c5d4bb17 | 25 | #include <asm/uaccess.h> |
4ff83ce1 | 26 | #include "as-layout.h" |
ba180fd4 | 27 | #include "kern_util.h" |
1da177e4 | 28 | #include "os.h" |
77bf4400 | 29 | #include "skas.h" |
1da177e4 | 30 | |
ba180fd4 JD |
31 | /* |
32 | * This is a per-cpu array. A processor only modifies its entry and it only | |
1da177e4 LT |
33 | * cares about its entry, so it's OK if another processor is modifying its |
34 | * entry. | |
35 | */ | |
36 | struct cpu_task cpu_tasks[NR_CPUS] = { [0 ... NR_CPUS - 1] = { -1, NULL } }; | |
37 | ||
2dc5802a | 38 | static inline int external_pid(void) |
1da177e4 | 39 | { |
77bf4400 | 40 | /* FIXME: Need to look up userspace_pid by cpu */ |
ba180fd4 | 41 | return userspace_pid[0]; |
1da177e4 LT |
42 | } |
43 | ||
44 | int pid_to_processor_id(int pid) | |
45 | { | |
46 | int i; | |
47 | ||
c5d4bb17 | 48 | for (i = 0; i < ncpus; i++) { |
ba180fd4 | 49 | if (cpu_tasks[i].pid == pid) |
6e21aec3 | 50 | return i; |
1da177e4 | 51 | } |
6e21aec3 | 52 | return -1; |
1da177e4 LT |
53 | } |
54 | ||
55 | void free_stack(unsigned long stack, int order) | |
56 | { | |
57 | free_pages(stack, order); | |
58 | } | |
59 | ||
60 | unsigned long alloc_stack(int order, int atomic) | |
61 | { | |
62 | unsigned long page; | |
53f9fc93 | 63 | gfp_t flags = GFP_KERNEL; |
1da177e4 | 64 | |
46db4a42 PBG |
65 | if (atomic) |
66 | flags = GFP_ATOMIC; | |
1da177e4 | 67 | page = __get_free_pages(flags, order); |
5c8aacea | 68 | |
6e21aec3 | 69 | return page; |
1da177e4 LT |
70 | } |
71 | ||
72 | int kernel_thread(int (*fn)(void *), void * arg, unsigned long flags) | |
73 | { | |
74 | int pid; | |
75 | ||
76 | current->thread.request.u.thread.proc = fn; | |
77 | current->thread.request.u.thread.arg = arg; | |
e0877f07 JD |
78 | pid = do_fork(CLONE_VM | CLONE_UNTRACED | flags, 0, |
79 | ¤t->thread.regs, 0, NULL, NULL); | |
6e21aec3 | 80 | return pid; |
1da177e4 | 81 | } |
73395a00 | 82 | EXPORT_SYMBOL(kernel_thread); |
1da177e4 | 83 | |
6e21aec3 | 84 | static inline void set_current(struct task_struct *task) |
1da177e4 | 85 | { |
ca9bc0bb | 86 | cpu_tasks[task_thread_info(task)->cpu] = ((struct cpu_task) |
2dc5802a | 87 | { external_pid(), task }); |
1da177e4 LT |
88 | } |
89 | ||
291248fd | 90 | extern void arch_switch_to(struct task_struct *to); |
77bf4400 | 91 | |
76b278ed | 92 | void *__switch_to(struct task_struct *from, struct task_struct *to) |
1da177e4 | 93 | { |
995473ae JD |
94 | to->thread.prev_sched = from; |
95 | set_current(to); | |
f6e34c6a | 96 | |
3eddddcf | 97 | do { |
6aa802ce | 98 | current->thread.saved_task = NULL; |
77bf4400 | 99 | |
c5d4bb17 JD |
100 | switch_threads(&from->thread.switch_buf, |
101 | &to->thread.switch_buf); | |
77bf4400 | 102 | |
291248fd | 103 | arch_switch_to(current); |
77bf4400 | 104 | |
ba180fd4 | 105 | if (current->thread.saved_task) |
3eddddcf | 106 | show_regs(&(current->thread.regs)); |
c5d4bb17 JD |
107 | to = current->thread.saved_task; |
108 | from = current; | |
291248fd | 109 | } while (current->thread.saved_task); |
f6e34c6a | 110 | |
6e21aec3 | 111 | return current->thread.prev_sched; |
1da177e4 LT |
112 | } |
113 | ||
114 | void interrupt_end(void) | |
115 | { | |
ba180fd4 | 116 | if (need_resched()) |
6e21aec3 | 117 | schedule(); |
d50349b0 | 118 | if (test_thread_flag(TIF_SIGPENDING)) |
6e21aec3 | 119 | do_signal(); |
d50349b0 AV |
120 | if (test_and_clear_thread_flag(TIF_NOTIFY_RESUME)) { |
121 | tracehook_notify_resume(¤t->thread.regs); | |
122 | if (current->replacement_session_keyring) | |
123 | key_replace_session_keyring(); | |
124 | } | |
1da177e4 LT |
125 | } |
126 | ||
1da177e4 LT |
127 | void exit_thread(void) |
128 | { | |
1da177e4 | 129 | } |
995473ae | 130 | |
c2220b2a | 131 | int get_current_pid(void) |
1da177e4 | 132 | { |
c2220b2a | 133 | return task_pid_nr(current); |
1da177e4 LT |
134 | } |
135 | ||
ba180fd4 JD |
136 | /* |
137 | * This is called magically, by its address being stuffed in a jmp_buf | |
77bf4400 JD |
138 | * and being longjmp-d to. |
139 | */ | |
140 | void new_thread_handler(void) | |
141 | { | |
142 | int (*fn)(void *), n; | |
143 | void *arg; | |
144 | ||
ba180fd4 | 145 | if (current->thread.prev_sched != NULL) |
77bf4400 JD |
146 | schedule_tail(current->thread.prev_sched); |
147 | current->thread.prev_sched = NULL; | |
148 | ||
149 | fn = current->thread.request.u.thread.proc; | |
150 | arg = current->thread.request.u.thread.arg; | |
151 | ||
ba180fd4 JD |
152 | /* |
153 | * The return value is 1 if the kernel thread execs a process, | |
77bf4400 JD |
154 | * 0 if it just exits |
155 | */ | |
156 | n = run_kernel_thread(fn, arg, ¤t->thread.exec_buf); | |
ba180fd4 | 157 | if (n == 1) { |
77bf4400 JD |
158 | /* Handle any immediate reschedules or signals */ |
159 | interrupt_end(); | |
160 | userspace(¤t->thread.regs.regs); | |
161 | } | |
162 | else do_exit(0); | |
163 | } | |
164 | ||
165 | /* Called magically, see new_thread_handler above */ | |
166 | void fork_handler(void) | |
167 | { | |
168 | force_flush_all(); | |
77bf4400 JD |
169 | |
170 | schedule_tail(current->thread.prev_sched); | |
171 | ||
ba180fd4 JD |
172 | /* |
173 | * XXX: if interrupt_end() calls schedule, this call to | |
77bf4400 | 174 | * arch_switch_to isn't needed. We could want to apply this to |
ba180fd4 JD |
175 | * improve performance. -bb |
176 | */ | |
291248fd | 177 | arch_switch_to(current); |
77bf4400 JD |
178 | |
179 | current->thread.prev_sched = NULL; | |
180 | ||
181 | /* Handle any immediate reschedules or signals */ | |
182 | interrupt_end(); | |
183 | ||
184 | userspace(¤t->thread.regs.regs); | |
185 | } | |
186 | ||
6f2c55b8 | 187 | int copy_thread(unsigned long clone_flags, unsigned long sp, |
995473ae | 188 | unsigned long stack_top, struct task_struct * p, |
1da177e4 LT |
189 | struct pt_regs *regs) |
190 | { | |
77bf4400 JD |
191 | void (*handler)(void); |
192 | int ret = 0; | |
aa6758d4 | 193 | |
1da177e4 | 194 | p->thread = (struct thread_struct) INIT_THREAD; |
aa6758d4 | 195 | |
ba180fd4 | 196 | if (current->thread.forking) { |
77bf4400 JD |
197 | memcpy(&p->thread.regs.regs, ®s->regs, |
198 | sizeof(p->thread.regs.regs)); | |
18baddda | 199 | REGS_SET_SYSCALL_RETURN(p->thread.regs.regs.gp, 0); |
ba180fd4 | 200 | if (sp != 0) |
18baddda | 201 | REGS_SP(p->thread.regs.regs.gp) = sp; |
aa6758d4 | 202 | |
77bf4400 | 203 | handler = fork_handler; |
aa6758d4 | 204 | |
77bf4400 JD |
205 | arch_copy_thread(¤t->thread.arch, &p->thread.arch); |
206 | } | |
207 | else { | |
fbfe9c84 | 208 | get_safe_registers(p->thread.regs.regs.gp, p->thread.regs.regs.fp); |
77bf4400 JD |
209 | p->thread.request.u.thread = current->thread.request.u.thread; |
210 | handler = new_thread_handler; | |
211 | } | |
212 | ||
213 | new_thread(task_stack_page(p), &p->thread.switch_buf, handler); | |
214 | ||
215 | if (current->thread.forking) { | |
216 | clear_flushed_tls(p); | |
217 | ||
218 | /* | |
219 | * Set a new TLS for the child thread? | |
220 | */ | |
221 | if (clone_flags & CLONE_SETTLS) | |
222 | ret = arch_copy_tls(p); | |
223 | } | |
aa6758d4 | 224 | |
aa6758d4 | 225 | return ret; |
1da177e4 LT |
226 | } |
227 | ||
228 | void initial_thread_cb(void (*proc)(void *), void *arg) | |
229 | { | |
230 | int save_kmalloc_ok = kmalloc_ok; | |
231 | ||
232 | kmalloc_ok = 0; | |
6aa802ce | 233 | initial_thread_cb_skas(proc, arg); |
1da177e4 LT |
234 | kmalloc_ok = save_kmalloc_ok; |
235 | } | |
995473ae | 236 | |
1da177e4 LT |
237 | void default_idle(void) |
238 | { | |
b160fb63 JD |
239 | unsigned long long nsecs; |
240 | ||
c5d4bb17 | 241 | while (1) { |
1da177e4 | 242 | /* endless idle loop with no priority at all */ |
1da177e4 LT |
243 | |
244 | /* | |
245 | * although we are an idle CPU, we do not want to | |
246 | * get into the scheduler unnecessarily. | |
247 | */ | |
ba180fd4 | 248 | if (need_resched()) |
1da177e4 | 249 | schedule(); |
995473ae | 250 | |
1268fbc7 FW |
251 | tick_nohz_idle_enter(); |
252 | rcu_idle_enter(); | |
b160fb63 JD |
253 | nsecs = disable_timer(); |
254 | idle_sleep(nsecs); | |
1268fbc7 FW |
255 | rcu_idle_exit(); |
256 | tick_nohz_idle_exit(); | |
1da177e4 LT |
257 | } |
258 | } | |
259 | ||
260 | void cpu_idle(void) | |
261 | { | |
a5a678c8 | 262 | cpu_tasks[current_thread_info()->cpu].pid = os_getpid(); |
77bf4400 | 263 | default_idle(); |
1da177e4 LT |
264 | } |
265 | ||
b6316293 PBG |
266 | int __cant_sleep(void) { |
267 | return in_atomic() || irqs_disabled() || in_interrupt(); | |
268 | /* Is in_interrupt() really needed? */ | |
1da177e4 LT |
269 | } |
270 | ||
1da177e4 LT |
271 | int user_context(unsigned long sp) |
272 | { | |
273 | unsigned long stack; | |
274 | ||
275 | stack = sp & (PAGE_MASK << CONFIG_KERNEL_STACK_ORDER); | |
a5a678c8 | 276 | return stack != (unsigned long) current_thread_info(); |
1da177e4 LT |
277 | } |
278 | ||
1da177e4 LT |
279 | extern exitcall_t __uml_exitcall_begin, __uml_exitcall_end; |
280 | ||
281 | void do_uml_exitcalls(void) | |
282 | { | |
283 | exitcall_t *call; | |
284 | ||
285 | call = &__uml_exitcall_end; | |
286 | while (--call >= &__uml_exitcall_begin) | |
287 | (*call)(); | |
288 | } | |
289 | ||
c0a9290e | 290 | char *uml_strdup(const char *string) |
1da177e4 | 291 | { |
dfe52244 | 292 | return kstrdup(string, GFP_KERNEL); |
1da177e4 | 293 | } |
73395a00 | 294 | EXPORT_SYMBOL(uml_strdup); |
1da177e4 | 295 | |
1da177e4 LT |
296 | int copy_to_user_proc(void __user *to, void *from, int size) |
297 | { | |
6e21aec3 | 298 | return copy_to_user(to, from, size); |
1da177e4 LT |
299 | } |
300 | ||
301 | int copy_from_user_proc(void *to, void __user *from, int size) | |
302 | { | |
6e21aec3 | 303 | return copy_from_user(to, from, size); |
1da177e4 LT |
304 | } |
305 | ||
306 | int clear_user_proc(void __user *buf, int size) | |
307 | { | |
6e21aec3 | 308 | return clear_user(buf, size); |
1da177e4 LT |
309 | } |
310 | ||
311 | int strlen_user_proc(char __user *str) | |
312 | { | |
6e21aec3 | 313 | return strlen_user(str); |
1da177e4 LT |
314 | } |
315 | ||
316 | int smp_sigio_handler(void) | |
317 | { | |
318 | #ifdef CONFIG_SMP | |
a5a678c8 | 319 | int cpu = current_thread_info()->cpu; |
1da177e4 | 320 | IPI_handler(cpu); |
ba180fd4 | 321 | if (cpu != 0) |
6e21aec3 | 322 | return 1; |
1da177e4 | 323 | #endif |
6e21aec3 | 324 | return 0; |
1da177e4 LT |
325 | } |
326 | ||
1da177e4 LT |
327 | int cpu(void) |
328 | { | |
a5a678c8 | 329 | return current_thread_info()->cpu; |
1da177e4 LT |
330 | } |
331 | ||
332 | static atomic_t using_sysemu = ATOMIC_INIT(0); | |
333 | int sysemu_supported; | |
334 | ||
335 | void set_using_sysemu(int value) | |
336 | { | |
337 | if (value > sysemu_supported) | |
338 | return; | |
339 | atomic_set(&using_sysemu, value); | |
340 | } | |
341 | ||
342 | int get_using_sysemu(void) | |
343 | { | |
344 | return atomic_read(&using_sysemu); | |
345 | } | |
346 | ||
6613c5e8 | 347 | static int sysemu_proc_show(struct seq_file *m, void *v) |
1da177e4 | 348 | { |
6613c5e8 AD |
349 | seq_printf(m, "%d\n", get_using_sysemu()); |
350 | return 0; | |
351 | } | |
1da177e4 | 352 | |
6613c5e8 AD |
353 | static int sysemu_proc_open(struct inode *inode, struct file *file) |
354 | { | |
355 | return single_open(file, sysemu_proc_show, NULL); | |
1da177e4 LT |
356 | } |
357 | ||
6613c5e8 AD |
358 | static ssize_t sysemu_proc_write(struct file *file, const char __user *buf, |
359 | size_t count, loff_t *pos) | |
1da177e4 LT |
360 | { |
361 | char tmp[2]; | |
362 | ||
363 | if (copy_from_user(tmp, buf, 1)) | |
364 | return -EFAULT; | |
365 | ||
366 | if (tmp[0] >= '0' && tmp[0] <= '2') | |
367 | set_using_sysemu(tmp[0] - '0'); | |
ba180fd4 JD |
368 | /* We use the first char, but pretend to write everything */ |
369 | return count; | |
1da177e4 LT |
370 | } |
371 | ||
6613c5e8 AD |
372 | static const struct file_operations sysemu_proc_fops = { |
373 | .owner = THIS_MODULE, | |
374 | .open = sysemu_proc_open, | |
375 | .read = seq_read, | |
376 | .llseek = seq_lseek, | |
377 | .release = single_release, | |
378 | .write = sysemu_proc_write, | |
379 | }; | |
380 | ||
1da177e4 LT |
381 | int __init make_proc_sysemu(void) |
382 | { | |
383 | struct proc_dir_entry *ent; | |
384 | if (!sysemu_supported) | |
385 | return 0; | |
386 | ||
6613c5e8 | 387 | ent = proc_create("sysemu", 0600, NULL, &sysemu_proc_fops); |
1da177e4 LT |
388 | |
389 | if (ent == NULL) | |
390 | { | |
30f417c6 | 391 | printk(KERN_WARNING "Failed to register /proc/sysemu\n"); |
6e21aec3 | 392 | return 0; |
1da177e4 LT |
393 | } |
394 | ||
1da177e4 LT |
395 | return 0; |
396 | } | |
397 | ||
398 | late_initcall(make_proc_sysemu); | |
399 | ||
400 | int singlestepping(void * t) | |
401 | { | |
402 | struct task_struct *task = t ? t : current; | |
403 | ||
c5d4bb17 | 404 | if (!(task->ptrace & PT_DTRACE)) |
ba180fd4 | 405 | return 0; |
1da177e4 LT |
406 | |
407 | if (task->thread.singlestep_syscall) | |
ba180fd4 | 408 | return 1; |
1da177e4 LT |
409 | |
410 | return 2; | |
411 | } | |
412 | ||
b8bd0220 BS |
413 | /* |
414 | * Only x86 and x86_64 have an arch_align_stack(). | |
415 | * All other arches have "#define arch_align_stack(x) (x)" | |
416 | * in their asm/system.h | |
417 | * As this is included in UML from asm-um/system-generic.h, | |
418 | * we can use it to behave as the subarch does. | |
419 | */ | |
420 | #ifndef arch_align_stack | |
1da177e4 LT |
421 | unsigned long arch_align_stack(unsigned long sp) |
422 | { | |
8f80e946 | 423 | if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space) |
1da177e4 LT |
424 | sp -= get_random_int() % 8192; |
425 | return sp & ~0xf; | |
426 | } | |
b8bd0220 | 427 | #endif |
c1127465 JD |
428 | |
429 | unsigned long get_wchan(struct task_struct *p) | |
430 | { | |
431 | unsigned long stack_page, sp, ip; | |
432 | bool seen_sched = 0; | |
433 | ||
434 | if ((p == NULL) || (p == current) || (p->state == TASK_RUNNING)) | |
435 | return 0; | |
436 | ||
437 | stack_page = (unsigned long) task_stack_page(p); | |
438 | /* Bail if the process has no kernel stack for some reason */ | |
439 | if (stack_page == 0) | |
440 | return 0; | |
441 | ||
442 | sp = p->thread.switch_buf->JB_SP; | |
443 | /* | |
444 | * Bail if the stack pointer is below the bottom of the kernel | |
445 | * stack for some reason | |
446 | */ | |
447 | if (sp < stack_page) | |
448 | return 0; | |
449 | ||
450 | while (sp < stack_page + THREAD_SIZE) { | |
451 | ip = *((unsigned long *) sp); | |
452 | if (in_sched_functions(ip)) | |
453 | /* Ignore everything until we're above the scheduler */ | |
454 | seen_sched = 1; | |
455 | else if (kernel_text_address(ip) && seen_sched) | |
456 | return ip; | |
457 | ||
458 | sp += sizeof(unsigned long); | |
459 | } | |
460 | ||
461 | return 0; | |
462 | } | |
8192ab42 JD |
463 | |
464 | int elf_core_copy_fpregs(struct task_struct *t, elf_fpregset_t *fpu) | |
465 | { | |
466 | int cpu = current_thread_info()->cpu; | |
467 | ||
468 | return save_fp_registers(userspace_pid[cpu], (unsigned long *) fpu); | |
469 | } | |
470 |