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> | |
10 | #include <linux/gfp.h> | |
11 | #include <linux/mm.h> | |
12 | #include <linux/personality.h> | |
13 | #include <linux/proc_fs.h> | |
14 | #include <linux/ptrace.h> | |
15 | #include <linux/random.h> | |
16 | #include <linux/sched.h> | |
17 | #include <linux/tick.h> | |
18 | #include <linux/threads.h> | |
19 | #include <asm/current.h> | |
20 | #include <asm/pgtable.h> | |
21 | #include <asm/uaccess.h> | |
4ff83ce1 | 22 | #include "as-layout.h" |
ba180fd4 | 23 | #include "kern_util.h" |
1da177e4 | 24 | #include "os.h" |
77bf4400 | 25 | #include "skas.h" |
ba180fd4 | 26 | #include "tlb.h" |
1da177e4 | 27 | |
ba180fd4 JD |
28 | /* |
29 | * This is a per-cpu array. A processor only modifies its entry and it only | |
1da177e4 LT |
30 | * cares about its entry, so it's OK if another processor is modifying its |
31 | * entry. | |
32 | */ | |
33 | struct cpu_task cpu_tasks[NR_CPUS] = { [0 ... NR_CPUS - 1] = { -1, NULL } }; | |
34 | ||
2dc5802a | 35 | static inline int external_pid(void) |
1da177e4 | 36 | { |
77bf4400 | 37 | /* FIXME: Need to look up userspace_pid by cpu */ |
ba180fd4 | 38 | return userspace_pid[0]; |
1da177e4 LT |
39 | } |
40 | ||
41 | int pid_to_processor_id(int pid) | |
42 | { | |
43 | int i; | |
44 | ||
c5d4bb17 | 45 | for (i = 0; i < ncpus; i++) { |
ba180fd4 | 46 | if (cpu_tasks[i].pid == pid) |
6e21aec3 | 47 | return i; |
1da177e4 | 48 | } |
6e21aec3 | 49 | return -1; |
1da177e4 LT |
50 | } |
51 | ||
52 | void free_stack(unsigned long stack, int order) | |
53 | { | |
54 | free_pages(stack, order); | |
55 | } | |
56 | ||
57 | unsigned long alloc_stack(int order, int atomic) | |
58 | { | |
59 | unsigned long page; | |
53f9fc93 | 60 | gfp_t flags = GFP_KERNEL; |
1da177e4 | 61 | |
46db4a42 PBG |
62 | if (atomic) |
63 | flags = GFP_ATOMIC; | |
1da177e4 | 64 | page = __get_free_pages(flags, order); |
5c8aacea | 65 | |
6e21aec3 | 66 | return page; |
1da177e4 LT |
67 | } |
68 | ||
69 | int kernel_thread(int (*fn)(void *), void * arg, unsigned long flags) | |
70 | { | |
71 | int pid; | |
72 | ||
73 | current->thread.request.u.thread.proc = fn; | |
74 | current->thread.request.u.thread.arg = arg; | |
e0877f07 JD |
75 | pid = do_fork(CLONE_VM | CLONE_UNTRACED | flags, 0, |
76 | ¤t->thread.regs, 0, NULL, NULL); | |
6e21aec3 | 77 | return pid; |
1da177e4 LT |
78 | } |
79 | ||
6e21aec3 | 80 | static inline void set_current(struct task_struct *task) |
1da177e4 | 81 | { |
ca9bc0bb | 82 | cpu_tasks[task_thread_info(task)->cpu] = ((struct cpu_task) |
2dc5802a | 83 | { external_pid(), task }); |
1da177e4 LT |
84 | } |
85 | ||
291248fd | 86 | extern void arch_switch_to(struct task_struct *to); |
77bf4400 | 87 | |
1da177e4 LT |
88 | void *_switch_to(void *prev, void *next, void *last) |
89 | { | |
995473ae | 90 | struct task_struct *from = prev; |
291248fd | 91 | struct task_struct *to = next; |
f6e34c6a | 92 | |
995473ae JD |
93 | to->thread.prev_sched = from; |
94 | set_current(to); | |
f6e34c6a | 95 | |
3eddddcf | 96 | do { |
6aa802ce | 97 | current->thread.saved_task = NULL; |
77bf4400 | 98 | |
c5d4bb17 JD |
99 | switch_threads(&from->thread.switch_buf, |
100 | &to->thread.switch_buf); | |
77bf4400 | 101 | |
291248fd | 102 | arch_switch_to(current); |
77bf4400 | 103 | |
ba180fd4 | 104 | if (current->thread.saved_task) |
3eddddcf | 105 | show_regs(&(current->thread.regs)); |
c5d4bb17 JD |
106 | to = current->thread.saved_task; |
107 | from = current; | |
291248fd | 108 | } while (current->thread.saved_task); |
f6e34c6a | 109 | |
6e21aec3 | 110 | return current->thread.prev_sched; |
f6e34c6a | 111 | |
1da177e4 LT |
112 | } |
113 | ||
114 | void interrupt_end(void) | |
115 | { | |
ba180fd4 | 116 | if (need_resched()) |
6e21aec3 | 117 | schedule(); |
ba180fd4 | 118 | if (test_tsk_thread_flag(current, TIF_SIGPENDING)) |
6e21aec3 | 119 | do_signal(); |
1da177e4 LT |
120 | } |
121 | ||
1da177e4 LT |
122 | void exit_thread(void) |
123 | { | |
1da177e4 | 124 | } |
995473ae | 125 | |
1da177e4 LT |
126 | void *get_current(void) |
127 | { | |
6e21aec3 | 128 | return current; |
1da177e4 LT |
129 | } |
130 | ||
ba180fd4 JD |
131 | /* |
132 | * This is called magically, by its address being stuffed in a jmp_buf | |
77bf4400 JD |
133 | * and being longjmp-d to. |
134 | */ | |
135 | void new_thread_handler(void) | |
136 | { | |
137 | int (*fn)(void *), n; | |
138 | void *arg; | |
139 | ||
ba180fd4 | 140 | if (current->thread.prev_sched != NULL) |
77bf4400 JD |
141 | schedule_tail(current->thread.prev_sched); |
142 | current->thread.prev_sched = NULL; | |
143 | ||
144 | fn = current->thread.request.u.thread.proc; | |
145 | arg = current->thread.request.u.thread.arg; | |
146 | ||
ba180fd4 JD |
147 | /* |
148 | * The return value is 1 if the kernel thread execs a process, | |
77bf4400 JD |
149 | * 0 if it just exits |
150 | */ | |
151 | n = run_kernel_thread(fn, arg, ¤t->thread.exec_buf); | |
ba180fd4 | 152 | if (n == 1) { |
77bf4400 JD |
153 | /* Handle any immediate reschedules or signals */ |
154 | interrupt_end(); | |
155 | userspace(¤t->thread.regs.regs); | |
156 | } | |
157 | else do_exit(0); | |
158 | } | |
159 | ||
160 | /* Called magically, see new_thread_handler above */ | |
161 | void fork_handler(void) | |
162 | { | |
163 | force_flush_all(); | |
77bf4400 JD |
164 | |
165 | schedule_tail(current->thread.prev_sched); | |
166 | ||
ba180fd4 JD |
167 | /* |
168 | * XXX: if interrupt_end() calls schedule, this call to | |
77bf4400 | 169 | * arch_switch_to isn't needed. We could want to apply this to |
ba180fd4 JD |
170 | * improve performance. -bb |
171 | */ | |
291248fd | 172 | arch_switch_to(current); |
77bf4400 JD |
173 | |
174 | current->thread.prev_sched = NULL; | |
175 | ||
176 | /* Handle any immediate reschedules or signals */ | |
177 | interrupt_end(); | |
178 | ||
179 | userspace(¤t->thread.regs.regs); | |
180 | } | |
181 | ||
6f2c55b8 | 182 | int copy_thread(unsigned long clone_flags, unsigned long sp, |
995473ae | 183 | unsigned long stack_top, struct task_struct * p, |
1da177e4 LT |
184 | struct pt_regs *regs) |
185 | { | |
77bf4400 JD |
186 | void (*handler)(void); |
187 | int ret = 0; | |
aa6758d4 | 188 | |
1da177e4 | 189 | p->thread = (struct thread_struct) INIT_THREAD; |
aa6758d4 | 190 | |
ba180fd4 | 191 | if (current->thread.forking) { |
77bf4400 JD |
192 | memcpy(&p->thread.regs.regs, ®s->regs, |
193 | sizeof(p->thread.regs.regs)); | |
18baddda | 194 | REGS_SET_SYSCALL_RETURN(p->thread.regs.regs.gp, 0); |
ba180fd4 | 195 | if (sp != 0) |
18baddda | 196 | REGS_SP(p->thread.regs.regs.gp) = sp; |
aa6758d4 | 197 | |
77bf4400 | 198 | handler = fork_handler; |
aa6758d4 | 199 | |
77bf4400 JD |
200 | arch_copy_thread(¤t->thread.arch, &p->thread.arch); |
201 | } | |
202 | else { | |
d25f2e12 | 203 | get_safe_registers(p->thread.regs.regs.gp); |
77bf4400 JD |
204 | p->thread.request.u.thread = current->thread.request.u.thread; |
205 | handler = new_thread_handler; | |
206 | } | |
207 | ||
208 | new_thread(task_stack_page(p), &p->thread.switch_buf, handler); | |
209 | ||
210 | if (current->thread.forking) { | |
211 | clear_flushed_tls(p); | |
212 | ||
213 | /* | |
214 | * Set a new TLS for the child thread? | |
215 | */ | |
216 | if (clone_flags & CLONE_SETTLS) | |
217 | ret = arch_copy_tls(p); | |
218 | } | |
aa6758d4 | 219 | |
aa6758d4 | 220 | return ret; |
1da177e4 LT |
221 | } |
222 | ||
223 | void initial_thread_cb(void (*proc)(void *), void *arg) | |
224 | { | |
225 | int save_kmalloc_ok = kmalloc_ok; | |
226 | ||
227 | kmalloc_ok = 0; | |
6aa802ce | 228 | initial_thread_cb_skas(proc, arg); |
1da177e4 LT |
229 | kmalloc_ok = save_kmalloc_ok; |
230 | } | |
995473ae | 231 | |
1da177e4 LT |
232 | void default_idle(void) |
233 | { | |
b160fb63 JD |
234 | unsigned long long nsecs; |
235 | ||
c5d4bb17 | 236 | while (1) { |
1da177e4 | 237 | /* endless idle loop with no priority at all */ |
1da177e4 LT |
238 | |
239 | /* | |
240 | * although we are an idle CPU, we do not want to | |
241 | * get into the scheduler unnecessarily. | |
242 | */ | |
ba180fd4 | 243 | if (need_resched()) |
1da177e4 | 244 | schedule(); |
995473ae | 245 | |
b8f8c3cf | 246 | tick_nohz_stop_sched_tick(1); |
b160fb63 JD |
247 | nsecs = disable_timer(); |
248 | idle_sleep(nsecs); | |
d2753a6d | 249 | tick_nohz_restart_sched_tick(); |
1da177e4 LT |
250 | } |
251 | } | |
252 | ||
253 | void cpu_idle(void) | |
254 | { | |
a5a678c8 | 255 | cpu_tasks[current_thread_info()->cpu].pid = os_getpid(); |
77bf4400 | 256 | default_idle(); |
1da177e4 LT |
257 | } |
258 | ||
b6316293 PBG |
259 | int __cant_sleep(void) { |
260 | return in_atomic() || irqs_disabled() || in_interrupt(); | |
261 | /* Is in_interrupt() really needed? */ | |
1da177e4 LT |
262 | } |
263 | ||
1da177e4 LT |
264 | int user_context(unsigned long sp) |
265 | { | |
266 | unsigned long stack; | |
267 | ||
268 | stack = sp & (PAGE_MASK << CONFIG_KERNEL_STACK_ORDER); | |
a5a678c8 | 269 | return stack != (unsigned long) current_thread_info(); |
1da177e4 LT |
270 | } |
271 | ||
1da177e4 LT |
272 | extern exitcall_t __uml_exitcall_begin, __uml_exitcall_end; |
273 | ||
274 | void do_uml_exitcalls(void) | |
275 | { | |
276 | exitcall_t *call; | |
277 | ||
278 | call = &__uml_exitcall_end; | |
279 | while (--call >= &__uml_exitcall_begin) | |
280 | (*call)(); | |
281 | } | |
282 | ||
c0a9290e | 283 | char *uml_strdup(const char *string) |
1da177e4 | 284 | { |
dfe52244 | 285 | return kstrdup(string, GFP_KERNEL); |
1da177e4 LT |
286 | } |
287 | ||
1da177e4 LT |
288 | int copy_to_user_proc(void __user *to, void *from, int size) |
289 | { | |
6e21aec3 | 290 | return copy_to_user(to, from, size); |
1da177e4 LT |
291 | } |
292 | ||
293 | int copy_from_user_proc(void *to, void __user *from, int size) | |
294 | { | |
6e21aec3 | 295 | return copy_from_user(to, from, size); |
1da177e4 LT |
296 | } |
297 | ||
298 | int clear_user_proc(void __user *buf, int size) | |
299 | { | |
6e21aec3 | 300 | return clear_user(buf, size); |
1da177e4 LT |
301 | } |
302 | ||
303 | int strlen_user_proc(char __user *str) | |
304 | { | |
6e21aec3 | 305 | return strlen_user(str); |
1da177e4 LT |
306 | } |
307 | ||
308 | int smp_sigio_handler(void) | |
309 | { | |
310 | #ifdef CONFIG_SMP | |
a5a678c8 | 311 | int cpu = current_thread_info()->cpu; |
1da177e4 | 312 | IPI_handler(cpu); |
ba180fd4 | 313 | if (cpu != 0) |
6e21aec3 | 314 | return 1; |
1da177e4 | 315 | #endif |
6e21aec3 | 316 | return 0; |
1da177e4 LT |
317 | } |
318 | ||
1da177e4 LT |
319 | int cpu(void) |
320 | { | |
a5a678c8 | 321 | return current_thread_info()->cpu; |
1da177e4 LT |
322 | } |
323 | ||
324 | static atomic_t using_sysemu = ATOMIC_INIT(0); | |
325 | int sysemu_supported; | |
326 | ||
327 | void set_using_sysemu(int value) | |
328 | { | |
329 | if (value > sysemu_supported) | |
330 | return; | |
331 | atomic_set(&using_sysemu, value); | |
332 | } | |
333 | ||
334 | int get_using_sysemu(void) | |
335 | { | |
336 | return atomic_read(&using_sysemu); | |
337 | } | |
338 | ||
339 | static int proc_read_sysemu(char *buf, char **start, off_t offset, int size,int *eof, void *data) | |
340 | { | |
ba180fd4 JD |
341 | if (snprintf(buf, size, "%d\n", get_using_sysemu()) < size) |
342 | /* No overflow */ | |
1da177e4 LT |
343 | *eof = 1; |
344 | ||
345 | return strlen(buf); | |
346 | } | |
347 | ||
4d338e1a | 348 | static int proc_write_sysemu(struct file *file,const char __user *buf, unsigned long count,void *data) |
1da177e4 LT |
349 | { |
350 | char tmp[2]; | |
351 | ||
352 | if (copy_from_user(tmp, buf, 1)) | |
353 | return -EFAULT; | |
354 | ||
355 | if (tmp[0] >= '0' && tmp[0] <= '2') | |
356 | set_using_sysemu(tmp[0] - '0'); | |
ba180fd4 JD |
357 | /* We use the first char, but pretend to write everything */ |
358 | return count; | |
1da177e4 LT |
359 | } |
360 | ||
361 | int __init make_proc_sysemu(void) | |
362 | { | |
363 | struct proc_dir_entry *ent; | |
364 | if (!sysemu_supported) | |
365 | return 0; | |
366 | ||
c74c120a | 367 | ent = create_proc_entry("sysemu", 0600, NULL); |
1da177e4 LT |
368 | |
369 | if (ent == NULL) | |
370 | { | |
30f417c6 | 371 | printk(KERN_WARNING "Failed to register /proc/sysemu\n"); |
6e21aec3 | 372 | return 0; |
1da177e4 LT |
373 | } |
374 | ||
375 | ent->read_proc = proc_read_sysemu; | |
376 | ent->write_proc = proc_write_sysemu; | |
377 | ||
378 | return 0; | |
379 | } | |
380 | ||
381 | late_initcall(make_proc_sysemu); | |
382 | ||
383 | int singlestepping(void * t) | |
384 | { | |
385 | struct task_struct *task = t ? t : current; | |
386 | ||
c5d4bb17 | 387 | if (!(task->ptrace & PT_DTRACE)) |
ba180fd4 | 388 | return 0; |
1da177e4 LT |
389 | |
390 | if (task->thread.singlestep_syscall) | |
ba180fd4 | 391 | return 1; |
1da177e4 LT |
392 | |
393 | return 2; | |
394 | } | |
395 | ||
b8bd0220 BS |
396 | /* |
397 | * Only x86 and x86_64 have an arch_align_stack(). | |
398 | * All other arches have "#define arch_align_stack(x) (x)" | |
399 | * in their asm/system.h | |
400 | * As this is included in UML from asm-um/system-generic.h, | |
401 | * we can use it to behave as the subarch does. | |
402 | */ | |
403 | #ifndef arch_align_stack | |
1da177e4 LT |
404 | unsigned long arch_align_stack(unsigned long sp) |
405 | { | |
8f80e946 | 406 | if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space) |
1da177e4 LT |
407 | sp -= get_random_int() % 8192; |
408 | return sp & ~0xf; | |
409 | } | |
b8bd0220 | 410 | #endif |
c1127465 JD |
411 | |
412 | unsigned long get_wchan(struct task_struct *p) | |
413 | { | |
414 | unsigned long stack_page, sp, ip; | |
415 | bool seen_sched = 0; | |
416 | ||
417 | if ((p == NULL) || (p == current) || (p->state == TASK_RUNNING)) | |
418 | return 0; | |
419 | ||
420 | stack_page = (unsigned long) task_stack_page(p); | |
421 | /* Bail if the process has no kernel stack for some reason */ | |
422 | if (stack_page == 0) | |
423 | return 0; | |
424 | ||
425 | sp = p->thread.switch_buf->JB_SP; | |
426 | /* | |
427 | * Bail if the stack pointer is below the bottom of the kernel | |
428 | * stack for some reason | |
429 | */ | |
430 | if (sp < stack_page) | |
431 | return 0; | |
432 | ||
433 | while (sp < stack_page + THREAD_SIZE) { | |
434 | ip = *((unsigned long *) sp); | |
435 | if (in_sched_functions(ip)) | |
436 | /* Ignore everything until we're above the scheduler */ | |
437 | seen_sched = 1; | |
438 | else if (kernel_text_address(ip) && seen_sched) | |
439 | return ip; | |
440 | ||
441 | sp += sizeof(unsigned long); | |
442 | } | |
443 | ||
444 | return 0; | |
445 | } | |
8192ab42 JD |
446 | |
447 | int elf_core_copy_fpregs(struct task_struct *t, elf_fpregset_t *fpu) | |
448 | { | |
449 | int cpu = current_thread_info()->cpu; | |
450 | ||
451 | return save_fp_registers(userspace_pid[cpu], (unsigned long *) fpu); | |
452 | } | |
453 |