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 | ||
ba180fd4 JD |
7 | #include "linux/stddef.h" |
8 | #include "linux/err.h" | |
9 | #include "linux/hardirq.h" | |
1da177e4 | 10 | #include "linux/mm.h" |
ba180fd4 | 11 | #include "linux/personality.h" |
1da177e4 LT |
12 | #include "linux/proc_fs.h" |
13 | #include "linux/ptrace.h" | |
14 | #include "linux/random.h" | |
ba180fd4 | 15 | #include "linux/sched.h" |
d2753a6d | 16 | #include "linux/tick.h" |
ba180fd4 | 17 | #include "linux/threads.h" |
1da177e4 | 18 | #include "asm/pgtable.h" |
1da177e4 | 19 | #include "asm/uaccess.h" |
4ff83ce1 | 20 | #include "as-layout.h" |
ba180fd4 | 21 | #include "kern_util.h" |
1da177e4 | 22 | #include "os.h" |
77bf4400 | 23 | #include "skas.h" |
ba180fd4 | 24 | #include "tlb.h" |
1da177e4 | 25 | |
ba180fd4 JD |
26 | /* |
27 | * This is a per-cpu array. A processor only modifies its entry and it only | |
1da177e4 LT |
28 | * cares about its entry, so it's OK if another processor is modifying its |
29 | * entry. | |
30 | */ | |
31 | struct cpu_task cpu_tasks[NR_CPUS] = { [0 ... NR_CPUS - 1] = { -1, NULL } }; | |
32 | ||
6e21aec3 | 33 | static inline int external_pid(struct task_struct *task) |
1da177e4 | 34 | { |
77bf4400 | 35 | /* FIXME: Need to look up userspace_pid by cpu */ |
ba180fd4 | 36 | return userspace_pid[0]; |
1da177e4 LT |
37 | } |
38 | ||
39 | int pid_to_processor_id(int pid) | |
40 | { | |
41 | int i; | |
42 | ||
ba180fd4 JD |
43 | for(i = 0; i < ncpus; i++) { |
44 | if (cpu_tasks[i].pid == pid) | |
6e21aec3 | 45 | return i; |
1da177e4 | 46 | } |
6e21aec3 | 47 | return -1; |
1da177e4 LT |
48 | } |
49 | ||
50 | void free_stack(unsigned long stack, int order) | |
51 | { | |
52 | free_pages(stack, order); | |
53 | } | |
54 | ||
55 | unsigned long alloc_stack(int order, int atomic) | |
56 | { | |
57 | unsigned long page; | |
53f9fc93 | 58 | gfp_t flags = GFP_KERNEL; |
1da177e4 | 59 | |
46db4a42 PBG |
60 | if (atomic) |
61 | flags = GFP_ATOMIC; | |
1da177e4 | 62 | page = __get_free_pages(flags, order); |
5c8aacea | 63 | |
6e21aec3 | 64 | return page; |
1da177e4 LT |
65 | } |
66 | ||
67 | int kernel_thread(int (*fn)(void *), void * arg, unsigned long flags) | |
68 | { | |
69 | int pid; | |
70 | ||
71 | current->thread.request.u.thread.proc = fn; | |
72 | current->thread.request.u.thread.arg = arg; | |
e0877f07 JD |
73 | pid = do_fork(CLONE_VM | CLONE_UNTRACED | flags, 0, |
74 | ¤t->thread.regs, 0, NULL, NULL); | |
6e21aec3 | 75 | return pid; |
1da177e4 LT |
76 | } |
77 | ||
6e21aec3 | 78 | static inline void set_current(struct task_struct *task) |
1da177e4 | 79 | { |
ca9bc0bb | 80 | cpu_tasks[task_thread_info(task)->cpu] = ((struct cpu_task) |
1da177e4 LT |
81 | { external_pid(task), task }); |
82 | } | |
83 | ||
291248fd | 84 | extern void arch_switch_to(struct task_struct *to); |
77bf4400 | 85 | |
1da177e4 LT |
86 | void *_switch_to(void *prev, void *next, void *last) |
87 | { | |
995473ae | 88 | struct task_struct *from = prev; |
291248fd | 89 | struct task_struct *to = next; |
f6e34c6a | 90 | |
995473ae JD |
91 | to->thread.prev_sched = from; |
92 | set_current(to); | |
f6e34c6a | 93 | |
3eddddcf | 94 | do { |
6aa802ce | 95 | current->thread.saved_task = NULL; |
77bf4400 | 96 | |
291248fd | 97 | switch_threads(&from->thread.switch_buf, &to->thread.switch_buf); |
77bf4400 | 98 | |
291248fd | 99 | arch_switch_to(current); |
77bf4400 | 100 | |
ba180fd4 | 101 | if (current->thread.saved_task) |
3eddddcf | 102 | show_regs(&(current->thread.regs)); |
291248fd KS |
103 | next = current->thread.saved_task; |
104 | prev = current; | |
105 | } while (current->thread.saved_task); | |
f6e34c6a | 106 | |
6e21aec3 | 107 | return current->thread.prev_sched; |
f6e34c6a | 108 | |
1da177e4 LT |
109 | } |
110 | ||
111 | void interrupt_end(void) | |
112 | { | |
ba180fd4 | 113 | if (need_resched()) |
6e21aec3 | 114 | schedule(); |
ba180fd4 | 115 | if (test_tsk_thread_flag(current, TIF_SIGPENDING)) |
6e21aec3 | 116 | do_signal(); |
1da177e4 LT |
117 | } |
118 | ||
1da177e4 LT |
119 | void exit_thread(void) |
120 | { | |
1da177e4 | 121 | } |
995473ae | 122 | |
1da177e4 LT |
123 | void *get_current(void) |
124 | { | |
6e21aec3 | 125 | return current; |
1da177e4 LT |
126 | } |
127 | ||
77bf4400 JD |
128 | extern void schedule_tail(struct task_struct *prev); |
129 | ||
ba180fd4 JD |
130 | /* |
131 | * This is called magically, by its address being stuffed in a jmp_buf | |
77bf4400 JD |
132 | * and being longjmp-d to. |
133 | */ | |
134 | void new_thread_handler(void) | |
135 | { | |
136 | int (*fn)(void *), n; | |
137 | void *arg; | |
138 | ||
ba180fd4 | 139 | if (current->thread.prev_sched != NULL) |
77bf4400 JD |
140 | schedule_tail(current->thread.prev_sched); |
141 | current->thread.prev_sched = NULL; | |
142 | ||
143 | fn = current->thread.request.u.thread.proc; | |
144 | arg = current->thread.request.u.thread.arg; | |
145 | ||
ba180fd4 JD |
146 | /* |
147 | * The return value is 1 if the kernel thread execs a process, | |
77bf4400 JD |
148 | * 0 if it just exits |
149 | */ | |
150 | n = run_kernel_thread(fn, arg, ¤t->thread.exec_buf); | |
ba180fd4 | 151 | if (n == 1) { |
77bf4400 JD |
152 | /* Handle any immediate reschedules or signals */ |
153 | interrupt_end(); | |
154 | userspace(¤t->thread.regs.regs); | |
155 | } | |
156 | else do_exit(0); | |
157 | } | |
158 | ||
159 | /* Called magically, see new_thread_handler above */ | |
160 | void fork_handler(void) | |
161 | { | |
162 | force_flush_all(); | |
77bf4400 JD |
163 | |
164 | schedule_tail(current->thread.prev_sched); | |
165 | ||
ba180fd4 JD |
166 | /* |
167 | * XXX: if interrupt_end() calls schedule, this call to | |
77bf4400 | 168 | * arch_switch_to isn't needed. We could want to apply this to |
ba180fd4 JD |
169 | * improve performance. -bb |
170 | */ | |
291248fd | 171 | arch_switch_to(current); |
77bf4400 JD |
172 | |
173 | current->thread.prev_sched = NULL; | |
174 | ||
175 | /* Handle any immediate reschedules or signals */ | |
176 | interrupt_end(); | |
177 | ||
178 | userspace(¤t->thread.regs.regs); | |
179 | } | |
180 | ||
1da177e4 | 181 | int copy_thread(int nr, unsigned long clone_flags, unsigned long sp, |
995473ae | 182 | unsigned long stack_top, struct task_struct * p, |
1da177e4 LT |
183 | struct pt_regs *regs) |
184 | { | |
77bf4400 JD |
185 | void (*handler)(void); |
186 | int ret = 0; | |
aa6758d4 | 187 | |
1da177e4 | 188 | p->thread = (struct thread_struct) INIT_THREAD; |
aa6758d4 | 189 | |
ba180fd4 | 190 | if (current->thread.forking) { |
77bf4400 JD |
191 | memcpy(&p->thread.regs.regs, ®s->regs, |
192 | sizeof(p->thread.regs.regs)); | |
18baddda | 193 | REGS_SET_SYSCALL_RETURN(p->thread.regs.regs.gp, 0); |
ba180fd4 | 194 | if (sp != 0) |
18baddda | 195 | REGS_SP(p->thread.regs.regs.gp) = sp; |
aa6758d4 | 196 | |
77bf4400 | 197 | handler = fork_handler; |
aa6758d4 | 198 | |
77bf4400 JD |
199 | arch_copy_thread(¤t->thread.arch, &p->thread.arch); |
200 | } | |
201 | else { | |
202 | init_thread_registers(&p->thread.regs.regs); | |
203 | p->thread.request.u.thread = current->thread.request.u.thread; | |
204 | handler = new_thread_handler; | |
205 | } | |
206 | ||
207 | new_thread(task_stack_page(p), &p->thread.switch_buf, handler); | |
208 | ||
209 | if (current->thread.forking) { | |
210 | clear_flushed_tls(p); | |
211 | ||
212 | /* | |
213 | * Set a new TLS for the child thread? | |
214 | */ | |
215 | if (clone_flags & CLONE_SETTLS) | |
216 | ret = arch_copy_tls(p); | |
217 | } | |
aa6758d4 | 218 | |
aa6758d4 | 219 | return ret; |
1da177e4 LT |
220 | } |
221 | ||
222 | void initial_thread_cb(void (*proc)(void *), void *arg) | |
223 | { | |
224 | int save_kmalloc_ok = kmalloc_ok; | |
225 | ||
226 | kmalloc_ok = 0; | |
6aa802ce | 227 | initial_thread_cb_skas(proc, arg); |
1da177e4 LT |
228 | kmalloc_ok = save_kmalloc_ok; |
229 | } | |
995473ae | 230 | |
1da177e4 LT |
231 | void default_idle(void) |
232 | { | |
b160fb63 JD |
233 | unsigned long long nsecs; |
234 | ||
ba180fd4 | 235 | while(1) { |
1da177e4 | 236 | /* endless idle loop with no priority at all */ |
1da177e4 LT |
237 | |
238 | /* | |
239 | * although we are an idle CPU, we do not want to | |
240 | * get into the scheduler unnecessarily. | |
241 | */ | |
ba180fd4 | 242 | if (need_resched()) |
1da177e4 | 243 | schedule(); |
995473ae | 244 | |
d2753a6d | 245 | tick_nohz_stop_sched_tick(); |
b160fb63 JD |
246 | nsecs = disable_timer(); |
247 | idle_sleep(nsecs); | |
d2753a6d | 248 | tick_nohz_restart_sched_tick(); |
1da177e4 LT |
249 | } |
250 | } | |
251 | ||
252 | void cpu_idle(void) | |
253 | { | |
77bf4400 JD |
254 | cpu_tasks[current_thread->cpu].pid = os_getpid(); |
255 | default_idle(); | |
1da177e4 LT |
256 | } |
257 | ||
1da177e4 LT |
258 | void dump_thread(struct pt_regs *regs, struct user *u) |
259 | { | |
260 | } | |
261 | ||
b6316293 PBG |
262 | int __cant_sleep(void) { |
263 | return in_atomic() || irqs_disabled() || in_interrupt(); | |
264 | /* Is in_interrupt() really needed? */ | |
1da177e4 LT |
265 | } |
266 | ||
1da177e4 LT |
267 | int user_context(unsigned long sp) |
268 | { | |
269 | unsigned long stack; | |
270 | ||
271 | stack = sp & (PAGE_MASK << CONFIG_KERNEL_STACK_ORDER); | |
6e21aec3 | 272 | return stack != (unsigned long) current_thread; |
1da177e4 LT |
273 | } |
274 | ||
1da177e4 LT |
275 | extern exitcall_t __uml_exitcall_begin, __uml_exitcall_end; |
276 | ||
277 | void do_uml_exitcalls(void) | |
278 | { | |
279 | exitcall_t *call; | |
280 | ||
281 | call = &__uml_exitcall_end; | |
282 | while (--call >= &__uml_exitcall_begin) | |
283 | (*call)(); | |
284 | } | |
285 | ||
c0a9290e | 286 | char *uml_strdup(const char *string) |
1da177e4 | 287 | { |
dfe52244 | 288 | return kstrdup(string, GFP_KERNEL); |
1da177e4 LT |
289 | } |
290 | ||
1da177e4 LT |
291 | int copy_to_user_proc(void __user *to, void *from, int size) |
292 | { | |
6e21aec3 | 293 | return copy_to_user(to, from, size); |
1da177e4 LT |
294 | } |
295 | ||
296 | int copy_from_user_proc(void *to, void __user *from, int size) | |
297 | { | |
6e21aec3 | 298 | return copy_from_user(to, from, size); |
1da177e4 LT |
299 | } |
300 | ||
301 | int clear_user_proc(void __user *buf, int size) | |
302 | { | |
6e21aec3 | 303 | return clear_user(buf, size); |
1da177e4 LT |
304 | } |
305 | ||
306 | int strlen_user_proc(char __user *str) | |
307 | { | |
6e21aec3 | 308 | return strlen_user(str); |
1da177e4 LT |
309 | } |
310 | ||
311 | int smp_sigio_handler(void) | |
312 | { | |
313 | #ifdef CONFIG_SMP | |
314 | int cpu = current_thread->cpu; | |
315 | IPI_handler(cpu); | |
ba180fd4 | 316 | if (cpu != 0) |
6e21aec3 | 317 | return 1; |
1da177e4 | 318 | #endif |
6e21aec3 | 319 | return 0; |
1da177e4 LT |
320 | } |
321 | ||
1da177e4 LT |
322 | int cpu(void) |
323 | { | |
6e21aec3 | 324 | return current_thread->cpu; |
1da177e4 LT |
325 | } |
326 | ||
327 | static atomic_t using_sysemu = ATOMIC_INIT(0); | |
328 | int sysemu_supported; | |
329 | ||
330 | void set_using_sysemu(int value) | |
331 | { | |
332 | if (value > sysemu_supported) | |
333 | return; | |
334 | atomic_set(&using_sysemu, value); | |
335 | } | |
336 | ||
337 | int get_using_sysemu(void) | |
338 | { | |
339 | return atomic_read(&using_sysemu); | |
340 | } | |
341 | ||
342 | static int proc_read_sysemu(char *buf, char **start, off_t offset, int size,int *eof, void *data) | |
343 | { | |
ba180fd4 JD |
344 | if (snprintf(buf, size, "%d\n", get_using_sysemu()) < size) |
345 | /* No overflow */ | |
1da177e4 LT |
346 | *eof = 1; |
347 | ||
348 | return strlen(buf); | |
349 | } | |
350 | ||
4d338e1a | 351 | static int proc_write_sysemu(struct file *file,const char __user *buf, unsigned long count,void *data) |
1da177e4 LT |
352 | { |
353 | char tmp[2]; | |
354 | ||
355 | if (copy_from_user(tmp, buf, 1)) | |
356 | return -EFAULT; | |
357 | ||
358 | if (tmp[0] >= '0' && tmp[0] <= '2') | |
359 | set_using_sysemu(tmp[0] - '0'); | |
ba180fd4 JD |
360 | /* We use the first char, but pretend to write everything */ |
361 | return count; | |
1da177e4 LT |
362 | } |
363 | ||
364 | int __init make_proc_sysemu(void) | |
365 | { | |
366 | struct proc_dir_entry *ent; | |
367 | if (!sysemu_supported) | |
368 | return 0; | |
369 | ||
370 | ent = create_proc_entry("sysemu", 0600, &proc_root); | |
371 | ||
372 | if (ent == NULL) | |
373 | { | |
30f417c6 | 374 | printk(KERN_WARNING "Failed to register /proc/sysemu\n"); |
6e21aec3 | 375 | return 0; |
1da177e4 LT |
376 | } |
377 | ||
378 | ent->read_proc = proc_read_sysemu; | |
379 | ent->write_proc = proc_write_sysemu; | |
380 | ||
381 | return 0; | |
382 | } | |
383 | ||
384 | late_initcall(make_proc_sysemu); | |
385 | ||
386 | int singlestepping(void * t) | |
387 | { | |
388 | struct task_struct *task = t ? t : current; | |
389 | ||
390 | if ( ! (task->ptrace & PT_DTRACE) ) | |
ba180fd4 | 391 | return 0; |
1da177e4 LT |
392 | |
393 | if (task->thread.singlestep_syscall) | |
ba180fd4 | 394 | return 1; |
1da177e4 LT |
395 | |
396 | return 2; | |
397 | } | |
398 | ||
b8bd0220 BS |
399 | /* |
400 | * Only x86 and x86_64 have an arch_align_stack(). | |
401 | * All other arches have "#define arch_align_stack(x) (x)" | |
402 | * in their asm/system.h | |
403 | * As this is included in UML from asm-um/system-generic.h, | |
404 | * we can use it to behave as the subarch does. | |
405 | */ | |
406 | #ifndef arch_align_stack | |
1da177e4 LT |
407 | unsigned long arch_align_stack(unsigned long sp) |
408 | { | |
8f80e946 | 409 | if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space) |
1da177e4 LT |
410 | sp -= get_random_int() % 8192; |
411 | return sp & ~0xf; | |
412 | } | |
b8bd0220 | 413 | #endif |
c1127465 JD |
414 | |
415 | unsigned long get_wchan(struct task_struct *p) | |
416 | { | |
417 | unsigned long stack_page, sp, ip; | |
418 | bool seen_sched = 0; | |
419 | ||
420 | if ((p == NULL) || (p == current) || (p->state == TASK_RUNNING)) | |
421 | return 0; | |
422 | ||
423 | stack_page = (unsigned long) task_stack_page(p); | |
424 | /* Bail if the process has no kernel stack for some reason */ | |
425 | if (stack_page == 0) | |
426 | return 0; | |
427 | ||
428 | sp = p->thread.switch_buf->JB_SP; | |
429 | /* | |
430 | * Bail if the stack pointer is below the bottom of the kernel | |
431 | * stack for some reason | |
432 | */ | |
433 | if (sp < stack_page) | |
434 | return 0; | |
435 | ||
436 | while (sp < stack_page + THREAD_SIZE) { | |
437 | ip = *((unsigned long *) sp); | |
438 | if (in_sched_functions(ip)) | |
439 | /* Ignore everything until we're above the scheduler */ | |
440 | seen_sched = 1; | |
441 | else if (kernel_text_address(ip) && seen_sched) | |
442 | return ip; | |
443 | ||
444 | sp += sizeof(unsigned long); | |
445 | } | |
446 | ||
447 | return 0; | |
448 | } |