]>
Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
1da177e4 LT |
2 | * Copyright (C) 1995 Linus Torvalds |
3 | * | |
4 | * Pentium III FXSR, SSE support | |
5 | * Gareth Hughes <gareth@valinux.com>, May 2000 | |
6612538c | 6 | * |
1da177e4 LT |
7 | * X86-64 port |
8 | * Andi Kleen. | |
76e4f660 AR |
9 | * |
10 | * CPU hotplug support - ashok.raj@intel.com | |
1da177e4 LT |
11 | */ |
12 | ||
13 | /* | |
14 | * This file handles the architecture-dependent parts of process handling.. | |
15 | */ | |
16 | ||
17 | #include <stdarg.h> | |
18 | ||
76e4f660 | 19 | #include <linux/cpu.h> |
1da177e4 LT |
20 | #include <linux/errno.h> |
21 | #include <linux/sched.h> | |
6612538c | 22 | #include <linux/fs.h> |
1da177e4 LT |
23 | #include <linux/kernel.h> |
24 | #include <linux/mm.h> | |
25 | #include <linux/elfcore.h> | |
26 | #include <linux/smp.h> | |
27 | #include <linux/slab.h> | |
28 | #include <linux/user.h> | |
1da177e4 LT |
29 | #include <linux/a.out.h> |
30 | #include <linux/interrupt.h> | |
6612538c | 31 | #include <linux/utsname.h> |
1da177e4 | 32 | #include <linux/delay.h> |
6612538c | 33 | #include <linux/module.h> |
1da177e4 | 34 | #include <linux/ptrace.h> |
1da177e4 | 35 | #include <linux/random.h> |
95833c83 | 36 | #include <linux/notifier.h> |
c6fd91f0 | 37 | #include <linux/kprobes.h> |
1eeb66a1 | 38 | #include <linux/kdebug.h> |
02290683 | 39 | #include <linux/tick.h> |
1da177e4 LT |
40 | |
41 | #include <asm/uaccess.h> | |
42 | #include <asm/pgtable.h> | |
43 | #include <asm/system.h> | |
44 | #include <asm/io.h> | |
45 | #include <asm/processor.h> | |
46 | #include <asm/i387.h> | |
47 | #include <asm/mmu_context.h> | |
48 | #include <asm/pda.h> | |
49 | #include <asm/prctl.h> | |
1da177e4 LT |
50 | #include <asm/desc.h> |
51 | #include <asm/proto.h> | |
52 | #include <asm/ia32.h> | |
95833c83 | 53 | #include <asm/idle.h> |
1da177e4 LT |
54 | |
55 | asmlinkage extern void ret_from_fork(void); | |
56 | ||
57 | unsigned long kernel_thread_flags = CLONE_VM | CLONE_UNTRACED; | |
58 | ||
1da177e4 LT |
59 | unsigned long boot_option_idle_override = 0; |
60 | EXPORT_SYMBOL(boot_option_idle_override); | |
61 | ||
62 | /* | |
63 | * Powermanagement idle function, if any.. | |
64 | */ | |
65 | void (*pm_idle)(void); | |
2ee60e17 | 66 | EXPORT_SYMBOL(pm_idle); |
1da177e4 LT |
67 | static DEFINE_PER_CPU(unsigned int, cpu_idle_state); |
68 | ||
e041c683 | 69 | static ATOMIC_NOTIFIER_HEAD(idle_notifier); |
95833c83 AK |
70 | |
71 | void idle_notifier_register(struct notifier_block *n) | |
72 | { | |
e041c683 | 73 | atomic_notifier_chain_register(&idle_notifier, n); |
95833c83 | 74 | } |
95833c83 | 75 | |
95833c83 AK |
76 | void enter_idle(void) |
77 | { | |
a15da49d | 78 | write_pda(isidle, 1); |
e041c683 | 79 | atomic_notifier_call_chain(&idle_notifier, IDLE_START, NULL); |
95833c83 AK |
80 | } |
81 | ||
82 | static void __exit_idle(void) | |
83 | { | |
9446868b | 84 | if (test_and_clear_bit_pda(0, isidle) == 0) |
a15da49d | 85 | return; |
e041c683 | 86 | atomic_notifier_call_chain(&idle_notifier, IDLE_END, NULL); |
95833c83 AK |
87 | } |
88 | ||
89 | /* Called from interrupts to signify idle end */ | |
90 | void exit_idle(void) | |
91 | { | |
a15da49d AK |
92 | /* idle loop has pid 0 */ |
93 | if (current->pid) | |
95833c83 AK |
94 | return; |
95 | __exit_idle(); | |
96 | } | |
97 | ||
1da177e4 LT |
98 | /* |
99 | * We use this if we don't have any better | |
100 | * idle routine.. | |
101 | */ | |
d8954222 | 102 | void default_idle(void) |
1da177e4 | 103 | { |
495ab9c0 | 104 | current_thread_info()->status &= ~TS_POLLING; |
0888f06a IM |
105 | /* |
106 | * TS_POLLING-cleared state must be visible before we | |
107 | * test NEED_RESCHED: | |
108 | */ | |
109 | smp_mb(); | |
72690a21 AK |
110 | local_irq_disable(); |
111 | if (!need_resched()) { | |
5ee613b6 IM |
112 | ktime_t t0, t1; |
113 | u64 t0n, t1n; | |
114 | ||
115 | t0 = ktime_get(); | |
116 | t0n = ktime_to_ns(t0); | |
117 | safe_halt(); /* enables interrupts racelessly */ | |
118 | local_irq_disable(); | |
119 | t1 = ktime_get(); | |
120 | t1n = ktime_to_ns(t1); | |
121 | sched_clock_idle_wakeup_event(t1n - t0n); | |
39d44a51 HS |
122 | } |
123 | local_irq_enable(); | |
495ab9c0 | 124 | current_thread_info()->status |= TS_POLLING; |
1da177e4 LT |
125 | } |
126 | ||
127 | /* | |
128 | * On SMP it's slightly faster (but much more power-consuming!) | |
129 | * to poll the ->need_resched flag instead of waiting for the | |
130 | * cross-CPU IPI to arrive. Use this option with caution. | |
131 | */ | |
6612538c | 132 | static void poll_idle(void) |
1da177e4 | 133 | { |
d331e739 | 134 | local_irq_enable(); |
72690a21 | 135 | cpu_relax(); |
1da177e4 LT |
136 | } |
137 | ||
76e4f660 AR |
138 | #ifdef CONFIG_HOTPLUG_CPU |
139 | DECLARE_PER_CPU(int, cpu_state); | |
140 | ||
141 | #include <asm/nmi.h> | |
1fa744e6 | 142 | /* We halt the CPU with physical CPU hotplug */ |
76e4f660 AR |
143 | static inline void play_dead(void) |
144 | { | |
145 | idle_task_exit(); | |
146 | wbinvd(); | |
147 | mb(); | |
148 | /* Ack it */ | |
149 | __get_cpu_var(cpu_state) = CPU_DEAD; | |
150 | ||
1fa744e6 | 151 | local_irq_disable(); |
76e4f660 | 152 | while (1) |
1fa744e6 | 153 | halt(); |
76e4f660 AR |
154 | } |
155 | #else | |
156 | static inline void play_dead(void) | |
157 | { | |
158 | BUG(); | |
159 | } | |
160 | #endif /* CONFIG_HOTPLUG_CPU */ | |
161 | ||
1da177e4 LT |
162 | /* |
163 | * The idle thread. There's no useful work to be | |
164 | * done, so just try to conserve power and have a | |
165 | * low exit latency (ie sit in a loop waiting for | |
166 | * somebody to say that they'd like to reschedule) | |
167 | */ | |
b10db7f0 | 168 | void cpu_idle(void) |
1da177e4 | 169 | { |
495ab9c0 | 170 | current_thread_info()->status |= TS_POLLING; |
1da177e4 LT |
171 | /* endless idle loop with no priority at all */ |
172 | while (1) { | |
173 | while (!need_resched()) { | |
174 | void (*idle)(void); | |
175 | ||
176 | if (__get_cpu_var(cpu_idle_state)) | |
177 | __get_cpu_var(cpu_idle_state) = 0; | |
178 | ||
02290683 CW |
179 | tick_nohz_stop_sched_tick(); |
180 | ||
1da177e4 LT |
181 | rmb(); |
182 | idle = pm_idle; | |
183 | if (!idle) | |
184 | idle = default_idle; | |
76e4f660 AR |
185 | if (cpu_is_offline(smp_processor_id())) |
186 | play_dead(); | |
d331e739 VP |
187 | /* |
188 | * Idle routines should keep interrupts disabled | |
189 | * from here on, until they go to idle. | |
190 | * Otherwise, idle callbacks can misfire. | |
191 | */ | |
192 | local_irq_disable(); | |
95833c83 | 193 | enter_idle(); |
1da177e4 | 194 | idle(); |
a15da49d AK |
195 | /* In many cases the interrupt that ended idle |
196 | has already called exit_idle. But some idle | |
197 | loops can be woken up without interrupt. */ | |
95833c83 | 198 | __exit_idle(); |
1da177e4 LT |
199 | } |
200 | ||
02290683 | 201 | tick_nohz_restart_sched_tick(); |
5bfb5d69 | 202 | preempt_enable_no_resched(); |
1da177e4 | 203 | schedule(); |
5bfb5d69 | 204 | preempt_disable(); |
1da177e4 LT |
205 | } |
206 | } | |
207 | ||
6612538c HS |
208 | static void do_nothing(void *unused) |
209 | { | |
210 | } | |
211 | ||
212 | void cpu_idle_wait(void) | |
213 | { | |
214 | unsigned int cpu, this_cpu = get_cpu(); | |
215 | cpumask_t map, tmp = current->cpus_allowed; | |
216 | ||
217 | set_cpus_allowed(current, cpumask_of_cpu(this_cpu)); | |
218 | put_cpu(); | |
219 | ||
220 | cpus_clear(map); | |
221 | for_each_online_cpu(cpu) { | |
222 | per_cpu(cpu_idle_state, cpu) = 1; | |
223 | cpu_set(cpu, map); | |
224 | } | |
225 | ||
226 | __get_cpu_var(cpu_idle_state) = 0; | |
227 | ||
228 | wmb(); | |
229 | do { | |
230 | ssleep(1); | |
231 | for_each_online_cpu(cpu) { | |
232 | if (cpu_isset(cpu, map) && !per_cpu(cpu_idle_state, cpu)) | |
233 | cpu_clear(cpu, map); | |
234 | } | |
235 | cpus_and(map, map, cpu_online_map); | |
236 | /* | |
237 | * We waited 1 sec, if a CPU still did not call idle | |
238 | * it may be because it is in idle and not waking up | |
239 | * because it has nothing to do. | |
240 | * Give all the remaining CPUS a kick. | |
241 | */ | |
242 | smp_call_function_mask(map, do_nothing, 0, 0); | |
243 | } while (!cpus_empty(map)); | |
244 | ||
245 | set_cpus_allowed(current, tmp); | |
246 | } | |
247 | EXPORT_SYMBOL_GPL(cpu_idle_wait); | |
248 | ||
1da177e4 LT |
249 | /* |
250 | * This uses new MONITOR/MWAIT instructions on P4 processors with PNI, | |
251 | * which can obviate IPI to trigger checking of need_resched. | |
252 | * We execute MONITOR against need_resched and enter optimized wait state | |
253 | * through MWAIT. Whenever someone changes need_resched, we would be woken | |
254 | * up from MWAIT (without an IPI). | |
991528d7 VP |
255 | * |
256 | * New with Core Duo processors, MWAIT can take some hints based on CPU | |
257 | * capability. | |
1da177e4 | 258 | */ |
65ea5b03 | 259 | void mwait_idle_with_hints(unsigned long ax, unsigned long cx) |
1da177e4 | 260 | { |
991528d7 | 261 | if (!need_resched()) { |
64c7c8f8 NP |
262 | __monitor((void *)¤t_thread_info()->flags, 0, 0); |
263 | smp_mb(); | |
991528d7 | 264 | if (!need_resched()) |
65ea5b03 | 265 | __mwait(ax, cx); |
1da177e4 LT |
266 | } |
267 | } | |
268 | ||
991528d7 VP |
269 | /* Default MONITOR/MWAIT with no hints, used for default C1 state */ |
270 | static void mwait_idle(void) | |
271 | { | |
d331e739 VP |
272 | if (!need_resched()) { |
273 | __monitor((void *)¤t_thread_info()->flags, 0, 0); | |
274 | smp_mb(); | |
275 | if (!need_resched()) | |
276 | __sti_mwait(0, 0); | |
277 | else | |
278 | local_irq_enable(); | |
279 | } else { | |
280 | local_irq_enable(); | |
281 | } | |
991528d7 VP |
282 | } |
283 | ||
e6982c67 | 284 | void __cpuinit select_idle_routine(const struct cpuinfo_x86 *c) |
1da177e4 LT |
285 | { |
286 | static int printed; | |
287 | if (cpu_has(c, X86_FEATURE_MWAIT)) { | |
288 | /* | |
289 | * Skip, if setup has overridden idle. | |
290 | * One CPU supports mwait => All CPUs supports mwait | |
291 | */ | |
292 | if (!pm_idle) { | |
293 | if (!printed) { | |
2d4fa2f6 | 294 | printk(KERN_INFO "using mwait in idle threads.\n"); |
1da177e4 LT |
295 | printed = 1; |
296 | } | |
297 | pm_idle = mwait_idle; | |
298 | } | |
299 | } | |
300 | } | |
301 | ||
6612538c | 302 | static int __init idle_setup(char *str) |
1da177e4 | 303 | { |
f039b754 | 304 | if (!strcmp(str, "poll")) { |
1da177e4 LT |
305 | printk("using polling idle threads.\n"); |
306 | pm_idle = poll_idle; | |
f039b754 AK |
307 | } else if (!strcmp(str, "mwait")) |
308 | force_mwait = 1; | |
309 | else | |
310 | return -1; | |
1da177e4 LT |
311 | |
312 | boot_option_idle_override = 1; | |
f039b754 | 313 | return 0; |
1da177e4 | 314 | } |
f039b754 | 315 | early_param("idle", idle_setup); |
1da177e4 | 316 | |
6612538c | 317 | /* Prints also some state that isn't saved in the pt_regs */ |
1da177e4 LT |
318 | void __show_regs(struct pt_regs * regs) |
319 | { | |
320 | unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L, fs, gs, shadowgs; | |
bb1995d5 | 321 | unsigned long d0, d1, d2, d3, d6, d7; |
6612538c HS |
322 | unsigned int fsindex, gsindex; |
323 | unsigned int ds, cs, es; | |
1da177e4 LT |
324 | |
325 | printk("\n"); | |
326 | print_modules(); | |
9acf23c4 AK |
327 | printk("Pid: %d, comm: %.20s %s %s %.*s\n", |
328 | current->pid, current->comm, print_tainted(), | |
96b644bd SH |
329 | init_utsname()->release, |
330 | (int)strcspn(init_utsname()->version, " "), | |
331 | init_utsname()->version); | |
65ea5b03 PA |
332 | printk("RIP: %04lx:[<%016lx>] ", regs->cs & 0xffff, regs->ip); |
333 | printk_address(regs->ip); | |
334 | printk("RSP: %04lx:%016lx EFLAGS: %08lx\n", regs->ss, regs->sp, | |
335 | regs->flags); | |
1da177e4 | 336 | printk("RAX: %016lx RBX: %016lx RCX: %016lx\n", |
65ea5b03 | 337 | regs->ax, regs->bx, regs->cx); |
1da177e4 | 338 | printk("RDX: %016lx RSI: %016lx RDI: %016lx\n", |
65ea5b03 | 339 | regs->dx, regs->si, regs->di); |
1da177e4 | 340 | printk("RBP: %016lx R08: %016lx R09: %016lx\n", |
65ea5b03 | 341 | regs->bp, regs->r8, regs->r9); |
1da177e4 LT |
342 | printk("R10: %016lx R11: %016lx R12: %016lx\n", |
343 | regs->r10, regs->r11, regs->r12); | |
344 | printk("R13: %016lx R14: %016lx R15: %016lx\n", | |
345 | regs->r13, regs->r14, regs->r15); | |
346 | ||
347 | asm("movl %%ds,%0" : "=r" (ds)); | |
348 | asm("movl %%cs,%0" : "=r" (cs)); | |
349 | asm("movl %%es,%0" : "=r" (es)); | |
350 | asm("movl %%fs,%0" : "=r" (fsindex)); | |
351 | asm("movl %%gs,%0" : "=r" (gsindex)); | |
352 | ||
353 | rdmsrl(MSR_FS_BASE, fs); | |
354 | rdmsrl(MSR_GS_BASE, gs); | |
355 | rdmsrl(MSR_KERNEL_GS_BASE, shadowgs); | |
356 | ||
f51c9452 GOC |
357 | cr0 = read_cr0(); |
358 | cr2 = read_cr2(); | |
359 | cr3 = read_cr3(); | |
360 | cr4 = read_cr4(); | |
1da177e4 LT |
361 | |
362 | printk("FS: %016lx(%04x) GS:%016lx(%04x) knlGS:%016lx\n", | |
363 | fs,fsindex,gs,gsindex,shadowgs); | |
364 | printk("CS: %04x DS: %04x ES: %04x CR0: %016lx\n", cs, ds, es, cr0); | |
365 | printk("CR2: %016lx CR3: %016lx CR4: %016lx\n", cr2, cr3, cr4); | |
bb1995d5 AS |
366 | |
367 | get_debugreg(d0, 0); | |
368 | get_debugreg(d1, 1); | |
369 | get_debugreg(d2, 2); | |
370 | printk("DR0: %016lx DR1: %016lx DR2: %016lx\n", d0, d1, d2); | |
371 | get_debugreg(d3, 3); | |
372 | get_debugreg(d6, 6); | |
373 | get_debugreg(d7, 7); | |
374 | printk("DR3: %016lx DR6: %016lx DR7: %016lx\n", d3, d6, d7); | |
1da177e4 LT |
375 | } |
376 | ||
377 | void show_regs(struct pt_regs *regs) | |
378 | { | |
c078d326 | 379 | printk("CPU %d:", smp_processor_id()); |
1da177e4 | 380 | __show_regs(regs); |
b538ed27 | 381 | show_trace(NULL, regs, (void *)(regs + 1)); |
1da177e4 LT |
382 | } |
383 | ||
384 | /* | |
385 | * Free current thread data structures etc.. | |
386 | */ | |
387 | void exit_thread(void) | |
388 | { | |
389 | struct task_struct *me = current; | |
390 | struct thread_struct *t = &me->thread; | |
73649dab | 391 | |
6612538c | 392 | if (me->thread.io_bitmap_ptr) { |
1da177e4 LT |
393 | struct tss_struct *tss = &per_cpu(init_tss, get_cpu()); |
394 | ||
395 | kfree(t->io_bitmap_ptr); | |
396 | t->io_bitmap_ptr = NULL; | |
d3a4f48d | 397 | clear_thread_flag(TIF_IO_BITMAP); |
1da177e4 LT |
398 | /* |
399 | * Careful, clear this in the TSS too: | |
400 | */ | |
401 | memset(tss->io_bitmap, 0xff, t->io_bitmap_max); | |
402 | t->io_bitmap_max = 0; | |
403 | put_cpu(); | |
404 | } | |
405 | } | |
406 | ||
407 | void flush_thread(void) | |
408 | { | |
409 | struct task_struct *tsk = current; | |
1da177e4 | 410 | |
303cd153 MD |
411 | if (test_tsk_thread_flag(tsk, TIF_ABI_PENDING)) { |
412 | clear_tsk_thread_flag(tsk, TIF_ABI_PENDING); | |
413 | if (test_tsk_thread_flag(tsk, TIF_IA32)) { | |
414 | clear_tsk_thread_flag(tsk, TIF_IA32); | |
415 | } else { | |
416 | set_tsk_thread_flag(tsk, TIF_IA32); | |
4d9bc79c | 417 | current_thread_info()->status |= TS_COMPAT; |
303cd153 | 418 | } |
4d9bc79c | 419 | } |
303cd153 | 420 | clear_tsk_thread_flag(tsk, TIF_DEBUG); |
1da177e4 LT |
421 | |
422 | tsk->thread.debugreg0 = 0; | |
423 | tsk->thread.debugreg1 = 0; | |
424 | tsk->thread.debugreg2 = 0; | |
425 | tsk->thread.debugreg3 = 0; | |
426 | tsk->thread.debugreg6 = 0; | |
427 | tsk->thread.debugreg7 = 0; | |
6612538c | 428 | memset(tsk->thread.tls_array, 0, sizeof(tsk->thread.tls_array)); |
1da177e4 LT |
429 | /* |
430 | * Forget coprocessor state.. | |
431 | */ | |
432 | clear_fpu(tsk); | |
433 | clear_used_math(); | |
434 | } | |
435 | ||
436 | void release_thread(struct task_struct *dead_task) | |
437 | { | |
438 | if (dead_task->mm) { | |
439 | if (dead_task->mm->context.size) { | |
440 | printk("WARNING: dead process %8s still has LDT? <%p/%d>\n", | |
441 | dead_task->comm, | |
442 | dead_task->mm->context.ldt, | |
443 | dead_task->mm->context.size); | |
444 | BUG(); | |
445 | } | |
446 | } | |
447 | } | |
448 | ||
449 | static inline void set_32bit_tls(struct task_struct *t, int tls, u32 addr) | |
450 | { | |
6612538c | 451 | struct user_desc ud = { |
1da177e4 LT |
452 | .base_addr = addr, |
453 | .limit = 0xfffff, | |
454 | .seg_32bit = 1, | |
455 | .limit_in_pages = 1, | |
456 | .useable = 1, | |
457 | }; | |
6842ef0e | 458 | struct desc_struct *desc = (void *)t->thread.tls_array; |
1da177e4 | 459 | desc += tls; |
80fbb69a | 460 | fill_ldt(desc, &ud); |
1da177e4 LT |
461 | } |
462 | ||
463 | static inline u32 read_32bit_tls(struct task_struct *t, int tls) | |
464 | { | |
91394eb0 | 465 | return get_desc_base(&t->thread.tls_array[tls]); |
1da177e4 LT |
466 | } |
467 | ||
468 | /* | |
469 | * This gets called before we allocate a new thread and copy | |
470 | * the current task into it. | |
471 | */ | |
472 | void prepare_to_copy(struct task_struct *tsk) | |
473 | { | |
474 | unlazy_fpu(tsk); | |
475 | } | |
476 | ||
65ea5b03 | 477 | int copy_thread(int nr, unsigned long clone_flags, unsigned long sp, |
1da177e4 LT |
478 | unsigned long unused, |
479 | struct task_struct * p, struct pt_regs * regs) | |
480 | { | |
481 | int err; | |
482 | struct pt_regs * childregs; | |
483 | struct task_struct *me = current; | |
484 | ||
a88cde13 | 485 | childregs = ((struct pt_regs *) |
57eafdc2 | 486 | (THREAD_SIZE + task_stack_page(p))) - 1; |
1da177e4 LT |
487 | *childregs = *regs; |
488 | ||
65ea5b03 PA |
489 | childregs->ax = 0; |
490 | childregs->sp = sp; | |
491 | if (sp == ~0UL) | |
492 | childregs->sp = (unsigned long)childregs; | |
1da177e4 | 493 | |
faca6227 PA |
494 | p->thread.sp = (unsigned long) childregs; |
495 | p->thread.sp0 = (unsigned long) (childregs+1); | |
496 | p->thread.usersp = me->thread.usersp; | |
1da177e4 | 497 | |
e4f17c43 | 498 | set_tsk_thread_flag(p, TIF_FORK); |
1da177e4 LT |
499 | |
500 | p->thread.fs = me->thread.fs; | |
501 | p->thread.gs = me->thread.gs; | |
502 | ||
fd51f666 L |
503 | asm("mov %%gs,%0" : "=m" (p->thread.gsindex)); |
504 | asm("mov %%fs,%0" : "=m" (p->thread.fsindex)); | |
505 | asm("mov %%es,%0" : "=m" (p->thread.es)); | |
506 | asm("mov %%ds,%0" : "=m" (p->thread.ds)); | |
1da177e4 | 507 | |
d3a4f48d | 508 | if (unlikely(test_tsk_thread_flag(me, TIF_IO_BITMAP))) { |
1da177e4 LT |
509 | p->thread.io_bitmap_ptr = kmalloc(IO_BITMAP_BYTES, GFP_KERNEL); |
510 | if (!p->thread.io_bitmap_ptr) { | |
511 | p->thread.io_bitmap_max = 0; | |
512 | return -ENOMEM; | |
513 | } | |
a88cde13 AK |
514 | memcpy(p->thread.io_bitmap_ptr, me->thread.io_bitmap_ptr, |
515 | IO_BITMAP_BYTES); | |
d3a4f48d | 516 | set_tsk_thread_flag(p, TIF_IO_BITMAP); |
6612538c | 517 | } |
1da177e4 LT |
518 | |
519 | /* | |
520 | * Set a new TLS for the child thread? | |
521 | */ | |
522 | if (clone_flags & CLONE_SETTLS) { | |
523 | #ifdef CONFIG_IA32_EMULATION | |
524 | if (test_thread_flag(TIF_IA32)) | |
efd1ca52 | 525 | err = do_set_thread_area(p, -1, |
65ea5b03 | 526 | (struct user_desc __user *)childregs->si, 0); |
1da177e4 LT |
527 | else |
528 | #endif | |
529 | err = do_arch_prctl(p, ARCH_SET_FS, childregs->r8); | |
530 | if (err) | |
531 | goto out; | |
532 | } | |
533 | err = 0; | |
534 | out: | |
535 | if (err && p->thread.io_bitmap_ptr) { | |
536 | kfree(p->thread.io_bitmap_ptr); | |
537 | p->thread.io_bitmap_max = 0; | |
538 | } | |
539 | return err; | |
540 | } | |
541 | ||
542 | /* | |
543 | * This special macro can be used to load a debugging register | |
544 | */ | |
6612538c HS |
545 | #define loaddebug(thread, r) set_debugreg(thread->debugreg ## r, r) |
546 | ||
547 | /* | |
548 | * Capture the user space registers if the task is not running (in user space) | |
549 | */ | |
550 | int dump_task_regs(struct task_struct *tsk, elf_gregset_t *regs) | |
551 | { | |
552 | struct pt_regs *pp, ptregs; | |
553 | ||
554 | pp = task_pt_regs(tsk); | |
555 | ||
556 | ptregs = *pp; | |
557 | ptregs.cs &= 0xffff; | |
558 | ptregs.ss &= 0xffff; | |
559 | ||
560 | elf_core_copy_regs(regs, &ptregs); | |
561 | ||
562 | return 1; | |
563 | } | |
1da177e4 | 564 | |
d3a4f48d | 565 | static inline void __switch_to_xtra(struct task_struct *prev_p, |
6612538c HS |
566 | struct task_struct *next_p, |
567 | struct tss_struct *tss) | |
d3a4f48d SE |
568 | { |
569 | struct thread_struct *prev, *next; | |
eee3af4a | 570 | unsigned long debugctl; |
d3a4f48d SE |
571 | |
572 | prev = &prev_p->thread, | |
573 | next = &next_p->thread; | |
574 | ||
eee3af4a MM |
575 | debugctl = prev->debugctlmsr; |
576 | if (next->ds_area_msr != prev->ds_area_msr) { | |
577 | /* we clear debugctl to make sure DS | |
578 | * is not in use when we change it */ | |
579 | debugctl = 0; | |
580 | wrmsrl(MSR_IA32_DEBUGCTLMSR, 0); | |
581 | wrmsrl(MSR_IA32_DS_AREA, next->ds_area_msr); | |
582 | } | |
583 | ||
584 | if (next->debugctlmsr != debugctl) | |
7e991604 RM |
585 | wrmsrl(MSR_IA32_DEBUGCTLMSR, next->debugctlmsr); |
586 | ||
d3a4f48d SE |
587 | if (test_tsk_thread_flag(next_p, TIF_DEBUG)) { |
588 | loaddebug(next, 0); | |
589 | loaddebug(next, 1); | |
590 | loaddebug(next, 2); | |
591 | loaddebug(next, 3); | |
592 | /* no 4 and 5 */ | |
593 | loaddebug(next, 6); | |
594 | loaddebug(next, 7); | |
595 | } | |
596 | ||
597 | if (test_tsk_thread_flag(next_p, TIF_IO_BITMAP)) { | |
598 | /* | |
599 | * Copy the relevant range of the IO bitmap. | |
600 | * Normally this is 128 bytes or less: | |
601 | */ | |
602 | memcpy(tss->io_bitmap, next->io_bitmap_ptr, | |
603 | max(prev->io_bitmap_max, next->io_bitmap_max)); | |
604 | } else if (test_tsk_thread_flag(prev_p, TIF_IO_BITMAP)) { | |
605 | /* | |
606 | * Clear any possible leftover bits: | |
607 | */ | |
608 | memset(tss->io_bitmap, 0xff, prev->io_bitmap_max); | |
609 | } | |
eee3af4a | 610 | |
eee3af4a MM |
611 | if (test_tsk_thread_flag(prev_p, TIF_BTS_TRACE_TS)) |
612 | ptrace_bts_take_timestamp(prev_p, BTS_TASK_DEPARTS); | |
613 | ||
614 | if (test_tsk_thread_flag(next_p, TIF_BTS_TRACE_TS)) | |
615 | ptrace_bts_take_timestamp(next_p, BTS_TASK_ARRIVES); | |
d3a4f48d SE |
616 | } |
617 | ||
1da177e4 LT |
618 | /* |
619 | * switch_to(x,y) should switch tasks from x to y. | |
620 | * | |
6612538c | 621 | * This could still be optimized: |
1da177e4 LT |
622 | * - fold all the options into a flag word and test it with a single test. |
623 | * - could test fs/gs bitsliced | |
099f318b AK |
624 | * |
625 | * Kprobes not supported here. Set the probe on schedule instead. | |
1da177e4 | 626 | */ |
f438d914 | 627 | struct task_struct * |
a88cde13 | 628 | __switch_to(struct task_struct *prev_p, struct task_struct *next_p) |
1da177e4 LT |
629 | { |
630 | struct thread_struct *prev = &prev_p->thread, | |
631 | *next = &next_p->thread; | |
6612538c | 632 | int cpu = smp_processor_id(); |
1da177e4 LT |
633 | struct tss_struct *tss = &per_cpu(init_tss, cpu); |
634 | ||
e07e23e1 AV |
635 | /* we're going to use this soon, after a few expensive things */ |
636 | if (next_p->fpu_counter>5) | |
637 | prefetch(&next->i387.fxsave); | |
638 | ||
1da177e4 LT |
639 | /* |
640 | * Reload esp0, LDT and the page table pointer: | |
641 | */ | |
ca241c75 | 642 | tss->x86_tss.sp0 = next->sp0; |
1da177e4 LT |
643 | |
644 | /* | |
645 | * Switch DS and ES. | |
646 | * This won't pick up thread selector changes, but I guess that is ok. | |
647 | */ | |
fd51f666 | 648 | asm volatile("mov %%es,%0" : "=m" (prev->es)); |
1da177e4 LT |
649 | if (unlikely(next->es | prev->es)) |
650 | loadsegment(es, next->es); | |
651 | ||
fd51f666 | 652 | asm volatile ("mov %%ds,%0" : "=m" (prev->ds)); |
1da177e4 LT |
653 | if (unlikely(next->ds | prev->ds)) |
654 | loadsegment(ds, next->ds); | |
655 | ||
656 | load_TLS(next, cpu); | |
657 | ||
658 | /* | |
659 | * Switch FS and GS. | |
660 | */ | |
661 | { | |
662 | unsigned fsindex; | |
663 | asm volatile("movl %%fs,%0" : "=r" (fsindex)); | |
664 | /* segment register != 0 always requires a reload. | |
665 | also reload when it has changed. | |
666 | when prev process used 64bit base always reload | |
667 | to avoid an information leak. */ | |
668 | if (unlikely(fsindex | next->fsindex | prev->fs)) { | |
669 | loadsegment(fs, next->fsindex); | |
670 | /* check if the user used a selector != 0 | |
671 | * if yes clear 64bit base, since overloaded base | |
672 | * is always mapped to the Null selector | |
673 | */ | |
674 | if (fsindex) | |
675 | prev->fs = 0; | |
676 | } | |
677 | /* when next process has a 64bit base use it */ | |
678 | if (next->fs) | |
679 | wrmsrl(MSR_FS_BASE, next->fs); | |
680 | prev->fsindex = fsindex; | |
681 | } | |
682 | { | |
683 | unsigned gsindex; | |
684 | asm volatile("movl %%gs,%0" : "=r" (gsindex)); | |
685 | if (unlikely(gsindex | next->gsindex | prev->gs)) { | |
686 | load_gs_index(next->gsindex); | |
687 | if (gsindex) | |
688 | prev->gs = 0; | |
689 | } | |
690 | if (next->gs) | |
691 | wrmsrl(MSR_KERNEL_GS_BASE, next->gs); | |
692 | prev->gsindex = gsindex; | |
693 | } | |
694 | ||
0a5ace2a AK |
695 | /* Must be after DS reload */ |
696 | unlazy_fpu(prev_p); | |
697 | ||
1da177e4 | 698 | /* |
45948d77 | 699 | * Switch the PDA and FPU contexts. |
1da177e4 | 700 | */ |
faca6227 PA |
701 | prev->usersp = read_pda(oldrsp); |
702 | write_pda(oldrsp, next->usersp); | |
1da177e4 | 703 | write_pda(pcurrent, next_p); |
18bd057b | 704 | |
a88cde13 | 705 | write_pda(kernelstack, |
7b0bda74 | 706 | (unsigned long)task_stack_page(next_p) + THREAD_SIZE - PDA_STACKOFFSET); |
0a425405 AV |
707 | #ifdef CONFIG_CC_STACKPROTECTOR |
708 | write_pda(stack_canary, next_p->stack_canary); | |
709 | /* | |
710 | * Build time only check to make sure the stack_canary is at | |
711 | * offset 40 in the pda; this is a gcc ABI requirement | |
712 | */ | |
713 | BUILD_BUG_ON(offsetof(struct x8664_pda, stack_canary) != 40); | |
714 | #endif | |
1da177e4 LT |
715 | |
716 | /* | |
d3a4f48d | 717 | * Now maybe reload the debug registers and handle I/O bitmaps |
1da177e4 | 718 | */ |
eee3af4a MM |
719 | if (unlikely(task_thread_info(next_p)->flags & _TIF_WORK_CTXSW_NEXT || |
720 | task_thread_info(prev_p)->flags & _TIF_WORK_CTXSW_PREV)) | |
d3a4f48d | 721 | __switch_to_xtra(prev_p, next_p, tss); |
1da177e4 | 722 | |
e07e23e1 AV |
723 | /* If the task has used fpu the last 5 timeslices, just do a full |
724 | * restore of the math state immediately to avoid the trap; the | |
725 | * chances of needing FPU soon are obviously high now | |
726 | */ | |
727 | if (next_p->fpu_counter>5) | |
728 | math_state_restore(); | |
1da177e4 LT |
729 | return prev_p; |
730 | } | |
731 | ||
732 | /* | |
733 | * sys_execve() executes a new program. | |
734 | */ | |
6612538c | 735 | asmlinkage |
1da177e4 LT |
736 | long sys_execve(char __user *name, char __user * __user *argv, |
737 | char __user * __user *envp, struct pt_regs regs) | |
738 | { | |
739 | long error; | |
740 | char * filename; | |
741 | ||
742 | filename = getname(name); | |
743 | error = PTR_ERR(filename); | |
744 | if (IS_ERR(filename)) | |
745 | return error; | |
746 | error = do_execve(filename, argv, envp, ®s); | |
1da177e4 LT |
747 | putname(filename); |
748 | return error; | |
749 | } | |
750 | ||
751 | void set_personality_64bit(void) | |
752 | { | |
753 | /* inherit personality from parent */ | |
754 | ||
755 | /* Make sure to be in 64bit mode */ | |
6612538c | 756 | clear_thread_flag(TIF_IA32); |
1da177e4 LT |
757 | |
758 | /* TBD: overwrites user setup. Should have two bits. | |
759 | But 64bit processes have always behaved this way, | |
760 | so it's not too bad. The main problem is just that | |
6612538c | 761 | 32bit childs are affected again. */ |
1da177e4 LT |
762 | current->personality &= ~READ_IMPLIES_EXEC; |
763 | } | |
764 | ||
765 | asmlinkage long sys_fork(struct pt_regs *regs) | |
766 | { | |
65ea5b03 | 767 | return do_fork(SIGCHLD, regs->sp, regs, 0, NULL, NULL); |
1da177e4 LT |
768 | } |
769 | ||
a88cde13 AK |
770 | asmlinkage long |
771 | sys_clone(unsigned long clone_flags, unsigned long newsp, | |
772 | void __user *parent_tid, void __user *child_tid, struct pt_regs *regs) | |
1da177e4 LT |
773 | { |
774 | if (!newsp) | |
65ea5b03 | 775 | newsp = regs->sp; |
1da177e4 LT |
776 | return do_fork(clone_flags, newsp, regs, 0, parent_tid, child_tid); |
777 | } | |
778 | ||
779 | /* | |
780 | * This is trivial, and on the face of it looks like it | |
781 | * could equally well be done in user mode. | |
782 | * | |
783 | * Not so, for quite unobvious reasons - register pressure. | |
784 | * In user mode vfork() cannot have a stack frame, and if | |
785 | * done by calling the "clone()" system call directly, you | |
786 | * do not have enough call-clobbered registers to hold all | |
787 | * the information you need. | |
788 | */ | |
789 | asmlinkage long sys_vfork(struct pt_regs *regs) | |
790 | { | |
65ea5b03 | 791 | return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, regs->sp, regs, 0, |
1da177e4 LT |
792 | NULL, NULL); |
793 | } | |
794 | ||
795 | unsigned long get_wchan(struct task_struct *p) | |
796 | { | |
797 | unsigned long stack; | |
65ea5b03 | 798 | u64 fp,ip; |
1da177e4 LT |
799 | int count = 0; |
800 | ||
801 | if (!p || p == current || p->state==TASK_RUNNING) | |
802 | return 0; | |
57eafdc2 | 803 | stack = (unsigned long)task_stack_page(p); |
faca6227 | 804 | if (p->thread.sp < stack || p->thread.sp > stack+THREAD_SIZE) |
1da177e4 | 805 | return 0; |
faca6227 | 806 | fp = *(u64 *)(p->thread.sp); |
1da177e4 | 807 | do { |
a88cde13 AK |
808 | if (fp < (unsigned long)stack || |
809 | fp > (unsigned long)stack+THREAD_SIZE) | |
1da177e4 | 810 | return 0; |
65ea5b03 PA |
811 | ip = *(u64 *)(fp+8); |
812 | if (!in_sched_functions(ip)) | |
813 | return ip; | |
1da177e4 LT |
814 | fp = *(u64 *)fp; |
815 | } while (count++ < 16); | |
816 | return 0; | |
817 | } | |
818 | ||
819 | long do_arch_prctl(struct task_struct *task, int code, unsigned long addr) | |
820 | { | |
821 | int ret = 0; | |
822 | int doit = task == current; | |
823 | int cpu; | |
824 | ||
825 | switch (code) { | |
826 | case ARCH_SET_GS: | |
84929801 | 827 | if (addr >= TASK_SIZE_OF(task)) |
1da177e4 LT |
828 | return -EPERM; |
829 | cpu = get_cpu(); | |
830 | /* handle small bases via the GDT because that's faster to | |
831 | switch. */ | |
832 | if (addr <= 0xffffffff) { | |
833 | set_32bit_tls(task, GS_TLS, addr); | |
834 | if (doit) { | |
835 | load_TLS(&task->thread, cpu); | |
836 | load_gs_index(GS_TLS_SEL); | |
837 | } | |
838 | task->thread.gsindex = GS_TLS_SEL; | |
839 | task->thread.gs = 0; | |
840 | } else { | |
841 | task->thread.gsindex = 0; | |
842 | task->thread.gs = addr; | |
843 | if (doit) { | |
a88cde13 AK |
844 | load_gs_index(0); |
845 | ret = checking_wrmsrl(MSR_KERNEL_GS_BASE, addr); | |
1da177e4 LT |
846 | } |
847 | } | |
848 | put_cpu(); | |
849 | break; | |
850 | case ARCH_SET_FS: | |
851 | /* Not strictly needed for fs, but do it for symmetry | |
852 | with gs */ | |
84929801 | 853 | if (addr >= TASK_SIZE_OF(task)) |
6612538c | 854 | return -EPERM; |
1da177e4 | 855 | cpu = get_cpu(); |
6612538c | 856 | /* handle small bases via the GDT because that's faster to |
1da177e4 | 857 | switch. */ |
6612538c | 858 | if (addr <= 0xffffffff) { |
1da177e4 | 859 | set_32bit_tls(task, FS_TLS, addr); |
6612538c HS |
860 | if (doit) { |
861 | load_TLS(&task->thread, cpu); | |
a88cde13 | 862 | asm volatile("movl %0,%%fs" :: "r"(FS_TLS_SEL)); |
1da177e4 LT |
863 | } |
864 | task->thread.fsindex = FS_TLS_SEL; | |
865 | task->thread.fs = 0; | |
6612538c | 866 | } else { |
1da177e4 LT |
867 | task->thread.fsindex = 0; |
868 | task->thread.fs = addr; | |
869 | if (doit) { | |
870 | /* set the selector to 0 to not confuse | |
871 | __switch_to */ | |
a88cde13 AK |
872 | asm volatile("movl %0,%%fs" :: "r" (0)); |
873 | ret = checking_wrmsrl(MSR_FS_BASE, addr); | |
1da177e4 LT |
874 | } |
875 | } | |
876 | put_cpu(); | |
877 | break; | |
6612538c HS |
878 | case ARCH_GET_FS: { |
879 | unsigned long base; | |
1da177e4 LT |
880 | if (task->thread.fsindex == FS_TLS_SEL) |
881 | base = read_32bit_tls(task, FS_TLS); | |
a88cde13 | 882 | else if (doit) |
1da177e4 | 883 | rdmsrl(MSR_FS_BASE, base); |
a88cde13 | 884 | else |
1da177e4 | 885 | base = task->thread.fs; |
6612538c HS |
886 | ret = put_user(base, (unsigned long __user *)addr); |
887 | break; | |
1da177e4 | 888 | } |
6612538c | 889 | case ARCH_GET_GS: { |
1da177e4 | 890 | unsigned long base; |
97c2803c | 891 | unsigned gsindex; |
1da177e4 LT |
892 | if (task->thread.gsindex == GS_TLS_SEL) |
893 | base = read_32bit_tls(task, GS_TLS); | |
97c2803c | 894 | else if (doit) { |
6612538c | 895 | asm("movl %%gs,%0" : "=r" (gsindex)); |
97c2803c JB |
896 | if (gsindex) |
897 | rdmsrl(MSR_KERNEL_GS_BASE, base); | |
898 | else | |
899 | base = task->thread.gs; | |
900 | } | |
a88cde13 | 901 | else |
1da177e4 | 902 | base = task->thread.gs; |
6612538c | 903 | ret = put_user(base, (unsigned long __user *)addr); |
1da177e4 LT |
904 | break; |
905 | } | |
906 | ||
907 | default: | |
908 | ret = -EINVAL; | |
909 | break; | |
6612538c | 910 | } |
1da177e4 | 911 | |
6612538c HS |
912 | return ret; |
913 | } | |
1da177e4 LT |
914 | |
915 | long sys_arch_prctl(int code, unsigned long addr) | |
916 | { | |
917 | return do_arch_prctl(current, code, addr); | |
1da177e4 LT |
918 | } |
919 | ||
920 | unsigned long arch_align_stack(unsigned long sp) | |
921 | { | |
c16b63e0 | 922 | if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space) |
1da177e4 LT |
923 | sp -= get_random_int() % 8192; |
924 | return sp & ~0xf; | |
925 | } | |
c1d171a0 JK |
926 | |
927 | unsigned long arch_randomize_brk(struct mm_struct *mm) | |
928 | { | |
929 | unsigned long range_end = mm->brk + 0x02000000; | |
930 | return randomize_range(mm->brk, range_end, 0) ? : mm->brk; | |
931 | } | |
932 |