]>
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 | |
6 | * | |
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> | |
22 | #include <linux/kernel.h> | |
23 | #include <linux/mm.h> | |
4e950f6f | 24 | #include <linux/fs.h> |
1da177e4 LT |
25 | #include <linux/elfcore.h> |
26 | #include <linux/smp.h> | |
27 | #include <linux/slab.h> | |
28 | #include <linux/user.h> | |
29 | #include <linux/module.h> | |
30 | #include <linux/a.out.h> | |
31 | #include <linux/interrupt.h> | |
32 | #include <linux/delay.h> | |
1da177e4 LT |
33 | #include <linux/ptrace.h> |
34 | #include <linux/utsname.h> | |
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 | */ | |
cdb04527 | 102 | static 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 | */ | |
132 | static void poll_idle (void) | |
133 | { | |
d331e739 | 134 | local_irq_enable(); |
72690a21 | 135 | cpu_relax(); |
1da177e4 LT |
136 | } |
137 | ||
40d6a146 SR |
138 | static void do_nothing(void *unused) |
139 | { | |
140 | } | |
141 | ||
1da177e4 LT |
142 | void cpu_idle_wait(void) |
143 | { | |
144 | unsigned int cpu, this_cpu = get_cpu(); | |
dc1829a4 | 145 | cpumask_t map, tmp = current->cpus_allowed; |
1da177e4 LT |
146 | |
147 | set_cpus_allowed(current, cpumask_of_cpu(this_cpu)); | |
148 | put_cpu(); | |
149 | ||
150 | cpus_clear(map); | |
151 | for_each_online_cpu(cpu) { | |
152 | per_cpu(cpu_idle_state, cpu) = 1; | |
153 | cpu_set(cpu, map); | |
154 | } | |
155 | ||
156 | __get_cpu_var(cpu_idle_state) = 0; | |
157 | ||
158 | wmb(); | |
159 | do { | |
160 | ssleep(1); | |
161 | for_each_online_cpu(cpu) { | |
a88cde13 AK |
162 | if (cpu_isset(cpu, map) && |
163 | !per_cpu(cpu_idle_state, cpu)) | |
1da177e4 LT |
164 | cpu_clear(cpu, map); |
165 | } | |
166 | cpus_and(map, map, cpu_online_map); | |
40d6a146 SR |
167 | /* |
168 | * We waited 1 sec, if a CPU still did not call idle | |
169 | * it may be because it is in idle and not waking up | |
170 | * because it has nothing to do. | |
171 | * Give all the remaining CPUS a kick. | |
172 | */ | |
173 | smp_call_function_mask(map, do_nothing, 0, 0); | |
1da177e4 | 174 | } while (!cpus_empty(map)); |
dc1829a4 IM |
175 | |
176 | set_cpus_allowed(current, tmp); | |
1da177e4 LT |
177 | } |
178 | EXPORT_SYMBOL_GPL(cpu_idle_wait); | |
179 | ||
76e4f660 AR |
180 | #ifdef CONFIG_HOTPLUG_CPU |
181 | DECLARE_PER_CPU(int, cpu_state); | |
182 | ||
183 | #include <asm/nmi.h> | |
1fa744e6 | 184 | /* We halt the CPU with physical CPU hotplug */ |
76e4f660 AR |
185 | static inline void play_dead(void) |
186 | { | |
187 | idle_task_exit(); | |
188 | wbinvd(); | |
189 | mb(); | |
190 | /* Ack it */ | |
191 | __get_cpu_var(cpu_state) = CPU_DEAD; | |
192 | ||
1fa744e6 | 193 | local_irq_disable(); |
76e4f660 | 194 | while (1) |
1fa744e6 | 195 | halt(); |
76e4f660 AR |
196 | } |
197 | #else | |
198 | static inline void play_dead(void) | |
199 | { | |
200 | BUG(); | |
201 | } | |
202 | #endif /* CONFIG_HOTPLUG_CPU */ | |
203 | ||
1da177e4 LT |
204 | /* |
205 | * The idle thread. There's no useful work to be | |
206 | * done, so just try to conserve power and have a | |
207 | * low exit latency (ie sit in a loop waiting for | |
208 | * somebody to say that they'd like to reschedule) | |
209 | */ | |
b10db7f0 | 210 | void cpu_idle(void) |
1da177e4 | 211 | { |
495ab9c0 | 212 | current_thread_info()->status |= TS_POLLING; |
1da177e4 LT |
213 | /* endless idle loop with no priority at all */ |
214 | while (1) { | |
215 | while (!need_resched()) { | |
216 | void (*idle)(void); | |
217 | ||
218 | if (__get_cpu_var(cpu_idle_state)) | |
219 | __get_cpu_var(cpu_idle_state) = 0; | |
220 | ||
02290683 CW |
221 | tick_nohz_stop_sched_tick(); |
222 | ||
1da177e4 LT |
223 | rmb(); |
224 | idle = pm_idle; | |
225 | if (!idle) | |
226 | idle = default_idle; | |
76e4f660 AR |
227 | if (cpu_is_offline(smp_processor_id())) |
228 | play_dead(); | |
d331e739 VP |
229 | /* |
230 | * Idle routines should keep interrupts disabled | |
231 | * from here on, until they go to idle. | |
232 | * Otherwise, idle callbacks can misfire. | |
233 | */ | |
234 | local_irq_disable(); | |
95833c83 | 235 | enter_idle(); |
1da177e4 | 236 | idle(); |
a15da49d AK |
237 | /* In many cases the interrupt that ended idle |
238 | has already called exit_idle. But some idle | |
239 | loops can be woken up without interrupt. */ | |
95833c83 | 240 | __exit_idle(); |
1da177e4 LT |
241 | } |
242 | ||
02290683 | 243 | tick_nohz_restart_sched_tick(); |
5bfb5d69 | 244 | preempt_enable_no_resched(); |
1da177e4 | 245 | schedule(); |
5bfb5d69 | 246 | preempt_disable(); |
1da177e4 LT |
247 | } |
248 | } | |
249 | ||
250 | /* | |
251 | * This uses new MONITOR/MWAIT instructions on P4 processors with PNI, | |
252 | * which can obviate IPI to trigger checking of need_resched. | |
253 | * We execute MONITOR against need_resched and enter optimized wait state | |
254 | * through MWAIT. Whenever someone changes need_resched, we would be woken | |
255 | * up from MWAIT (without an IPI). | |
991528d7 VP |
256 | * |
257 | * New with Core Duo processors, MWAIT can take some hints based on CPU | |
258 | * capability. | |
1da177e4 | 259 | */ |
991528d7 | 260 | void mwait_idle_with_hints(unsigned long eax, unsigned long ecx) |
1da177e4 | 261 | { |
991528d7 | 262 | if (!need_resched()) { |
64c7c8f8 NP |
263 | __monitor((void *)¤t_thread_info()->flags, 0, 0); |
264 | smp_mb(); | |
991528d7 VP |
265 | if (!need_resched()) |
266 | __mwait(eax, ecx); | |
1da177e4 LT |
267 | } |
268 | } | |
269 | ||
991528d7 VP |
270 | /* Default MONITOR/MWAIT with no hints, used for default C1 state */ |
271 | static void mwait_idle(void) | |
272 | { | |
d331e739 VP |
273 | if (!need_resched()) { |
274 | __monitor((void *)¤t_thread_info()->flags, 0, 0); | |
275 | smp_mb(); | |
276 | if (!need_resched()) | |
277 | __sti_mwait(0, 0); | |
278 | else | |
279 | local_irq_enable(); | |
280 | } else { | |
281 | local_irq_enable(); | |
282 | } | |
991528d7 VP |
283 | } |
284 | ||
e6982c67 | 285 | void __cpuinit select_idle_routine(const struct cpuinfo_x86 *c) |
1da177e4 LT |
286 | { |
287 | static int printed; | |
288 | if (cpu_has(c, X86_FEATURE_MWAIT)) { | |
289 | /* | |
290 | * Skip, if setup has overridden idle. | |
291 | * One CPU supports mwait => All CPUs supports mwait | |
292 | */ | |
293 | if (!pm_idle) { | |
294 | if (!printed) { | |
2d4fa2f6 | 295 | printk(KERN_INFO "using mwait in idle threads.\n"); |
1da177e4 LT |
296 | printed = 1; |
297 | } | |
298 | pm_idle = mwait_idle; | |
299 | } | |
300 | } | |
301 | } | |
302 | ||
303 | static int __init idle_setup (char *str) | |
304 | { | |
f039b754 | 305 | if (!strcmp(str, "poll")) { |
1da177e4 LT |
306 | printk("using polling idle threads.\n"); |
307 | pm_idle = poll_idle; | |
f039b754 AK |
308 | } else if (!strcmp(str, "mwait")) |
309 | force_mwait = 1; | |
310 | else | |
311 | return -1; | |
1da177e4 LT |
312 | |
313 | boot_option_idle_override = 1; | |
f039b754 | 314 | return 0; |
1da177e4 | 315 | } |
f039b754 | 316 | early_param("idle", idle_setup); |
1da177e4 LT |
317 | |
318 | /* Prints also some state that isn't saved in the pt_regs */ | |
319 | void __show_regs(struct pt_regs * regs) | |
320 | { | |
321 | unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L, fs, gs, shadowgs; | |
bb1995d5 | 322 | unsigned long d0, d1, d2, d3, d6, d7; |
1da177e4 LT |
323 | unsigned int fsindex,gsindex; |
324 | unsigned int ds,cs,es; | |
325 | ||
326 | printk("\n"); | |
327 | print_modules(); | |
9acf23c4 AK |
328 | printk("Pid: %d, comm: %.20s %s %s %.*s\n", |
329 | current->pid, current->comm, print_tainted(), | |
96b644bd SH |
330 | init_utsname()->release, |
331 | (int)strcspn(init_utsname()->version, " "), | |
332 | init_utsname()->version); | |
1da177e4 LT |
333 | printk("RIP: %04lx:[<%016lx>] ", regs->cs & 0xffff, regs->rip); |
334 | printk_address(regs->rip); | |
3ac94932 | 335 | printk("RSP: %04lx:%016lx EFLAGS: %08lx\n", regs->ss, regs->rsp, |
a88cde13 | 336 | regs->eflags); |
1da177e4 LT |
337 | printk("RAX: %016lx RBX: %016lx RCX: %016lx\n", |
338 | regs->rax, regs->rbx, regs->rcx); | |
339 | printk("RDX: %016lx RSI: %016lx RDI: %016lx\n", | |
340 | regs->rdx, regs->rsi, regs->rdi); | |
341 | printk("RBP: %016lx R08: %016lx R09: %016lx\n", | |
342 | regs->rbp, regs->r8, regs->r9); | |
343 | printk("R10: %016lx R11: %016lx R12: %016lx\n", | |
344 | regs->r10, regs->r11, regs->r12); | |
345 | printk("R13: %016lx R14: %016lx R15: %016lx\n", | |
346 | regs->r13, regs->r14, regs->r15); | |
347 | ||
348 | asm("movl %%ds,%0" : "=r" (ds)); | |
349 | asm("movl %%cs,%0" : "=r" (cs)); | |
350 | asm("movl %%es,%0" : "=r" (es)); | |
351 | asm("movl %%fs,%0" : "=r" (fsindex)); | |
352 | asm("movl %%gs,%0" : "=r" (gsindex)); | |
353 | ||
354 | rdmsrl(MSR_FS_BASE, fs); | |
355 | rdmsrl(MSR_GS_BASE, gs); | |
356 | rdmsrl(MSR_KERNEL_GS_BASE, shadowgs); | |
357 | ||
f51c9452 GOC |
358 | cr0 = read_cr0(); |
359 | cr2 = read_cr2(); | |
360 | cr3 = read_cr3(); | |
361 | cr4 = read_cr4(); | |
1da177e4 LT |
362 | |
363 | printk("FS: %016lx(%04x) GS:%016lx(%04x) knlGS:%016lx\n", | |
364 | fs,fsindex,gs,gsindex,shadowgs); | |
365 | printk("CS: %04x DS: %04x ES: %04x CR0: %016lx\n", cs, ds, es, cr0); | |
366 | printk("CR2: %016lx CR3: %016lx CR4: %016lx\n", cr2, cr3, cr4); | |
bb1995d5 AS |
367 | |
368 | get_debugreg(d0, 0); | |
369 | get_debugreg(d1, 1); | |
370 | get_debugreg(d2, 2); | |
371 | printk("DR0: %016lx DR1: %016lx DR2: %016lx\n", d0, d1, d2); | |
372 | get_debugreg(d3, 3); | |
373 | get_debugreg(d6, 6); | |
374 | get_debugreg(d7, 7); | |
375 | printk("DR3: %016lx DR6: %016lx DR7: %016lx\n", d3, d6, d7); | |
1da177e4 LT |
376 | } |
377 | ||
378 | void show_regs(struct pt_regs *regs) | |
379 | { | |
c078d326 | 380 | printk("CPU %d:", smp_processor_id()); |
1da177e4 | 381 | __show_regs(regs); |
b538ed27 | 382 | show_trace(NULL, regs, (void *)(regs + 1)); |
1da177e4 LT |
383 | } |
384 | ||
385 | /* | |
386 | * Free current thread data structures etc.. | |
387 | */ | |
388 | void exit_thread(void) | |
389 | { | |
390 | struct task_struct *me = current; | |
391 | struct thread_struct *t = &me->thread; | |
73649dab | 392 | |
1da177e4 LT |
393 | if (me->thread.io_bitmap_ptr) { |
394 | struct tss_struct *tss = &per_cpu(init_tss, get_cpu()); | |
395 | ||
396 | kfree(t->io_bitmap_ptr); | |
397 | t->io_bitmap_ptr = NULL; | |
d3a4f48d | 398 | clear_thread_flag(TIF_IO_BITMAP); |
1da177e4 LT |
399 | /* |
400 | * Careful, clear this in the TSS too: | |
401 | */ | |
402 | memset(tss->io_bitmap, 0xff, t->io_bitmap_max); | |
403 | t->io_bitmap_max = 0; | |
404 | put_cpu(); | |
405 | } | |
406 | } | |
407 | ||
408 | void flush_thread(void) | |
409 | { | |
410 | struct task_struct *tsk = current; | |
1da177e4 | 411 | |
303cd153 MD |
412 | if (test_tsk_thread_flag(tsk, TIF_ABI_PENDING)) { |
413 | clear_tsk_thread_flag(tsk, TIF_ABI_PENDING); | |
414 | if (test_tsk_thread_flag(tsk, TIF_IA32)) { | |
415 | clear_tsk_thread_flag(tsk, TIF_IA32); | |
416 | } else { | |
417 | set_tsk_thread_flag(tsk, TIF_IA32); | |
4d9bc79c | 418 | current_thread_info()->status |= TS_COMPAT; |
303cd153 | 419 | } |
4d9bc79c | 420 | } |
303cd153 | 421 | clear_tsk_thread_flag(tsk, TIF_DEBUG); |
1da177e4 LT |
422 | |
423 | tsk->thread.debugreg0 = 0; | |
424 | tsk->thread.debugreg1 = 0; | |
425 | tsk->thread.debugreg2 = 0; | |
426 | tsk->thread.debugreg3 = 0; | |
427 | tsk->thread.debugreg6 = 0; | |
428 | tsk->thread.debugreg7 = 0; | |
429 | memset(tsk->thread.tls_array, 0, sizeof(tsk->thread.tls_array)); | |
430 | /* | |
431 | * Forget coprocessor state.. | |
432 | */ | |
433 | clear_fpu(tsk); | |
434 | clear_used_math(); | |
435 | } | |
436 | ||
437 | void release_thread(struct task_struct *dead_task) | |
438 | { | |
439 | if (dead_task->mm) { | |
440 | if (dead_task->mm->context.size) { | |
441 | printk("WARNING: dead process %8s still has LDT? <%p/%d>\n", | |
442 | dead_task->comm, | |
443 | dead_task->mm->context.ldt, | |
444 | dead_task->mm->context.size); | |
445 | BUG(); | |
446 | } | |
447 | } | |
448 | } | |
449 | ||
450 | static inline void set_32bit_tls(struct task_struct *t, int tls, u32 addr) | |
451 | { | |
452 | struct user_desc ud = { | |
453 | .base_addr = addr, | |
454 | .limit = 0xfffff, | |
455 | .seg_32bit = 1, | |
456 | .limit_in_pages = 1, | |
457 | .useable = 1, | |
458 | }; | |
459 | struct n_desc_struct *desc = (void *)t->thread.tls_array; | |
460 | desc += tls; | |
461 | desc->a = LDT_entry_a(&ud); | |
462 | desc->b = LDT_entry_b(&ud); | |
463 | } | |
464 | ||
465 | static inline u32 read_32bit_tls(struct task_struct *t, int tls) | |
466 | { | |
91394eb0 | 467 | return get_desc_base(&t->thread.tls_array[tls]); |
1da177e4 LT |
468 | } |
469 | ||
470 | /* | |
471 | * This gets called before we allocate a new thread and copy | |
472 | * the current task into it. | |
473 | */ | |
474 | void prepare_to_copy(struct task_struct *tsk) | |
475 | { | |
476 | unlazy_fpu(tsk); | |
477 | } | |
478 | ||
479 | int copy_thread(int nr, unsigned long clone_flags, unsigned long rsp, | |
480 | unsigned long unused, | |
481 | struct task_struct * p, struct pt_regs * regs) | |
482 | { | |
483 | int err; | |
484 | struct pt_regs * childregs; | |
485 | struct task_struct *me = current; | |
486 | ||
a88cde13 | 487 | childregs = ((struct pt_regs *) |
57eafdc2 | 488 | (THREAD_SIZE + task_stack_page(p))) - 1; |
1da177e4 LT |
489 | *childregs = *regs; |
490 | ||
491 | childregs->rax = 0; | |
492 | childregs->rsp = rsp; | |
a88cde13 | 493 | if (rsp == ~0UL) |
1da177e4 | 494 | childregs->rsp = (unsigned long)childregs; |
1da177e4 LT |
495 | |
496 | p->thread.rsp = (unsigned long) childregs; | |
497 | p->thread.rsp0 = (unsigned long) (childregs+1); | |
498 | p->thread.userrsp = me->thread.userrsp; | |
499 | ||
e4f17c43 | 500 | set_tsk_thread_flag(p, TIF_FORK); |
1da177e4 LT |
501 | |
502 | p->thread.fs = me->thread.fs; | |
503 | p->thread.gs = me->thread.gs; | |
504 | ||
fd51f666 L |
505 | asm("mov %%gs,%0" : "=m" (p->thread.gsindex)); |
506 | asm("mov %%fs,%0" : "=m" (p->thread.fsindex)); | |
507 | asm("mov %%es,%0" : "=m" (p->thread.es)); | |
508 | asm("mov %%ds,%0" : "=m" (p->thread.ds)); | |
1da177e4 | 509 | |
d3a4f48d | 510 | if (unlikely(test_tsk_thread_flag(me, TIF_IO_BITMAP))) { |
1da177e4 LT |
511 | p->thread.io_bitmap_ptr = kmalloc(IO_BITMAP_BYTES, GFP_KERNEL); |
512 | if (!p->thread.io_bitmap_ptr) { | |
513 | p->thread.io_bitmap_max = 0; | |
514 | return -ENOMEM; | |
515 | } | |
a88cde13 AK |
516 | memcpy(p->thread.io_bitmap_ptr, me->thread.io_bitmap_ptr, |
517 | IO_BITMAP_BYTES); | |
d3a4f48d | 518 | set_tsk_thread_flag(p, TIF_IO_BITMAP); |
1da177e4 LT |
519 | } |
520 | ||
521 | /* | |
522 | * Set a new TLS for the child thread? | |
523 | */ | |
524 | if (clone_flags & CLONE_SETTLS) { | |
525 | #ifdef CONFIG_IA32_EMULATION | |
526 | if (test_thread_flag(TIF_IA32)) | |
527 | err = ia32_child_tls(p, childregs); | |
528 | else | |
529 | #endif | |
530 | err = do_arch_prctl(p, ARCH_SET_FS, childregs->r8); | |
531 | if (err) | |
532 | goto out; | |
533 | } | |
534 | err = 0; | |
535 | out: | |
536 | if (err && p->thread.io_bitmap_ptr) { | |
537 | kfree(p->thread.io_bitmap_ptr); | |
538 | p->thread.io_bitmap_max = 0; | |
539 | } | |
540 | return err; | |
541 | } | |
542 | ||
543 | /* | |
544 | * This special macro can be used to load a debugging register | |
545 | */ | |
2b514e74 | 546 | #define loaddebug(thread,r) set_debugreg(thread->debugreg ## r, r) |
1da177e4 | 547 | |
d3a4f48d SE |
548 | static inline void __switch_to_xtra(struct task_struct *prev_p, |
549 | struct task_struct *next_p, | |
550 | struct tss_struct *tss) | |
551 | { | |
552 | struct thread_struct *prev, *next; | |
553 | ||
554 | prev = &prev_p->thread, | |
555 | next = &next_p->thread; | |
556 | ||
557 | if (test_tsk_thread_flag(next_p, TIF_DEBUG)) { | |
558 | loaddebug(next, 0); | |
559 | loaddebug(next, 1); | |
560 | loaddebug(next, 2); | |
561 | loaddebug(next, 3); | |
562 | /* no 4 and 5 */ | |
563 | loaddebug(next, 6); | |
564 | loaddebug(next, 7); | |
565 | } | |
566 | ||
567 | if (test_tsk_thread_flag(next_p, TIF_IO_BITMAP)) { | |
568 | /* | |
569 | * Copy the relevant range of the IO bitmap. | |
570 | * Normally this is 128 bytes or less: | |
571 | */ | |
572 | memcpy(tss->io_bitmap, next->io_bitmap_ptr, | |
573 | max(prev->io_bitmap_max, next->io_bitmap_max)); | |
574 | } else if (test_tsk_thread_flag(prev_p, TIF_IO_BITMAP)) { | |
575 | /* | |
576 | * Clear any possible leftover bits: | |
577 | */ | |
578 | memset(tss->io_bitmap, 0xff, prev->io_bitmap_max); | |
579 | } | |
580 | } | |
581 | ||
1da177e4 LT |
582 | /* |
583 | * switch_to(x,y) should switch tasks from x to y. | |
584 | * | |
585 | * This could still be optimized: | |
586 | * - fold all the options into a flag word and test it with a single test. | |
587 | * - could test fs/gs bitsliced | |
099f318b AK |
588 | * |
589 | * Kprobes not supported here. Set the probe on schedule instead. | |
1da177e4 | 590 | */ |
f438d914 | 591 | struct task_struct * |
a88cde13 | 592 | __switch_to(struct task_struct *prev_p, struct task_struct *next_p) |
1da177e4 LT |
593 | { |
594 | struct thread_struct *prev = &prev_p->thread, | |
595 | *next = &next_p->thread; | |
596 | int cpu = smp_processor_id(); | |
597 | struct tss_struct *tss = &per_cpu(init_tss, cpu); | |
598 | ||
e07e23e1 AV |
599 | /* we're going to use this soon, after a few expensive things */ |
600 | if (next_p->fpu_counter>5) | |
601 | prefetch(&next->i387.fxsave); | |
602 | ||
1da177e4 LT |
603 | /* |
604 | * Reload esp0, LDT and the page table pointer: | |
605 | */ | |
606 | tss->rsp0 = next->rsp0; | |
607 | ||
608 | /* | |
609 | * Switch DS and ES. | |
610 | * This won't pick up thread selector changes, but I guess that is ok. | |
611 | */ | |
fd51f666 | 612 | asm volatile("mov %%es,%0" : "=m" (prev->es)); |
1da177e4 LT |
613 | if (unlikely(next->es | prev->es)) |
614 | loadsegment(es, next->es); | |
615 | ||
fd51f666 | 616 | asm volatile ("mov %%ds,%0" : "=m" (prev->ds)); |
1da177e4 LT |
617 | if (unlikely(next->ds | prev->ds)) |
618 | loadsegment(ds, next->ds); | |
619 | ||
620 | load_TLS(next, cpu); | |
621 | ||
622 | /* | |
623 | * Switch FS and GS. | |
624 | */ | |
625 | { | |
626 | unsigned fsindex; | |
627 | asm volatile("movl %%fs,%0" : "=r" (fsindex)); | |
628 | /* segment register != 0 always requires a reload. | |
629 | also reload when it has changed. | |
630 | when prev process used 64bit base always reload | |
631 | to avoid an information leak. */ | |
632 | if (unlikely(fsindex | next->fsindex | prev->fs)) { | |
633 | loadsegment(fs, next->fsindex); | |
634 | /* check if the user used a selector != 0 | |
635 | * if yes clear 64bit base, since overloaded base | |
636 | * is always mapped to the Null selector | |
637 | */ | |
638 | if (fsindex) | |
639 | prev->fs = 0; | |
640 | } | |
641 | /* when next process has a 64bit base use it */ | |
642 | if (next->fs) | |
643 | wrmsrl(MSR_FS_BASE, next->fs); | |
644 | prev->fsindex = fsindex; | |
645 | } | |
646 | { | |
647 | unsigned gsindex; | |
648 | asm volatile("movl %%gs,%0" : "=r" (gsindex)); | |
649 | if (unlikely(gsindex | next->gsindex | prev->gs)) { | |
650 | load_gs_index(next->gsindex); | |
651 | if (gsindex) | |
652 | prev->gs = 0; | |
653 | } | |
654 | if (next->gs) | |
655 | wrmsrl(MSR_KERNEL_GS_BASE, next->gs); | |
656 | prev->gsindex = gsindex; | |
657 | } | |
658 | ||
0a5ace2a AK |
659 | /* Must be after DS reload */ |
660 | unlazy_fpu(prev_p); | |
661 | ||
1da177e4 | 662 | /* |
45948d77 | 663 | * Switch the PDA and FPU contexts. |
1da177e4 LT |
664 | */ |
665 | prev->userrsp = read_pda(oldrsp); | |
666 | write_pda(oldrsp, next->userrsp); | |
667 | write_pda(pcurrent, next_p); | |
18bd057b | 668 | |
a88cde13 | 669 | write_pda(kernelstack, |
7b0bda74 | 670 | (unsigned long)task_stack_page(next_p) + THREAD_SIZE - PDA_STACKOFFSET); |
0a425405 AV |
671 | #ifdef CONFIG_CC_STACKPROTECTOR |
672 | write_pda(stack_canary, next_p->stack_canary); | |
673 | /* | |
674 | * Build time only check to make sure the stack_canary is at | |
675 | * offset 40 in the pda; this is a gcc ABI requirement | |
676 | */ | |
677 | BUILD_BUG_ON(offsetof(struct x8664_pda, stack_canary) != 40); | |
678 | #endif | |
1da177e4 LT |
679 | |
680 | /* | |
d3a4f48d | 681 | * Now maybe reload the debug registers and handle I/O bitmaps |
1da177e4 | 682 | */ |
d3a4f48d SE |
683 | if (unlikely((task_thread_info(next_p)->flags & _TIF_WORK_CTXSW)) |
684 | || test_tsk_thread_flag(prev_p, TIF_IO_BITMAP)) | |
685 | __switch_to_xtra(prev_p, next_p, tss); | |
1da177e4 | 686 | |
e07e23e1 AV |
687 | /* If the task has used fpu the last 5 timeslices, just do a full |
688 | * restore of the math state immediately to avoid the trap; the | |
689 | * chances of needing FPU soon are obviously high now | |
690 | */ | |
691 | if (next_p->fpu_counter>5) | |
692 | math_state_restore(); | |
1da177e4 LT |
693 | return prev_p; |
694 | } | |
695 | ||
696 | /* | |
697 | * sys_execve() executes a new program. | |
698 | */ | |
699 | asmlinkage | |
700 | long sys_execve(char __user *name, char __user * __user *argv, | |
701 | char __user * __user *envp, struct pt_regs regs) | |
702 | { | |
703 | long error; | |
704 | char * filename; | |
705 | ||
706 | filename = getname(name); | |
707 | error = PTR_ERR(filename); | |
708 | if (IS_ERR(filename)) | |
709 | return error; | |
710 | error = do_execve(filename, argv, envp, ®s); | |
711 | if (error == 0) { | |
712 | task_lock(current); | |
713 | current->ptrace &= ~PT_DTRACE; | |
714 | task_unlock(current); | |
715 | } | |
716 | putname(filename); | |
717 | return error; | |
718 | } | |
719 | ||
720 | void set_personality_64bit(void) | |
721 | { | |
722 | /* inherit personality from parent */ | |
723 | ||
724 | /* Make sure to be in 64bit mode */ | |
725 | clear_thread_flag(TIF_IA32); | |
726 | ||
727 | /* TBD: overwrites user setup. Should have two bits. | |
728 | But 64bit processes have always behaved this way, | |
729 | so it's not too bad. The main problem is just that | |
730 | 32bit childs are affected again. */ | |
731 | current->personality &= ~READ_IMPLIES_EXEC; | |
732 | } | |
733 | ||
734 | asmlinkage long sys_fork(struct pt_regs *regs) | |
735 | { | |
736 | return do_fork(SIGCHLD, regs->rsp, regs, 0, NULL, NULL); | |
737 | } | |
738 | ||
a88cde13 AK |
739 | asmlinkage long |
740 | sys_clone(unsigned long clone_flags, unsigned long newsp, | |
741 | void __user *parent_tid, void __user *child_tid, struct pt_regs *regs) | |
1da177e4 LT |
742 | { |
743 | if (!newsp) | |
744 | newsp = regs->rsp; | |
745 | return do_fork(clone_flags, newsp, regs, 0, parent_tid, child_tid); | |
746 | } | |
747 | ||
748 | /* | |
749 | * This is trivial, and on the face of it looks like it | |
750 | * could equally well be done in user mode. | |
751 | * | |
752 | * Not so, for quite unobvious reasons - register pressure. | |
753 | * In user mode vfork() cannot have a stack frame, and if | |
754 | * done by calling the "clone()" system call directly, you | |
755 | * do not have enough call-clobbered registers to hold all | |
756 | * the information you need. | |
757 | */ | |
758 | asmlinkage long sys_vfork(struct pt_regs *regs) | |
759 | { | |
760 | return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, regs->rsp, regs, 0, | |
761 | NULL, NULL); | |
762 | } | |
763 | ||
764 | unsigned long get_wchan(struct task_struct *p) | |
765 | { | |
766 | unsigned long stack; | |
767 | u64 fp,rip; | |
768 | int count = 0; | |
769 | ||
770 | if (!p || p == current || p->state==TASK_RUNNING) | |
771 | return 0; | |
57eafdc2 | 772 | stack = (unsigned long)task_stack_page(p); |
1da177e4 LT |
773 | if (p->thread.rsp < stack || p->thread.rsp > stack+THREAD_SIZE) |
774 | return 0; | |
775 | fp = *(u64 *)(p->thread.rsp); | |
776 | do { | |
a88cde13 AK |
777 | if (fp < (unsigned long)stack || |
778 | fp > (unsigned long)stack+THREAD_SIZE) | |
1da177e4 LT |
779 | return 0; |
780 | rip = *(u64 *)(fp+8); | |
781 | if (!in_sched_functions(rip)) | |
782 | return rip; | |
783 | fp = *(u64 *)fp; | |
784 | } while (count++ < 16); | |
785 | return 0; | |
786 | } | |
787 | ||
788 | long do_arch_prctl(struct task_struct *task, int code, unsigned long addr) | |
789 | { | |
790 | int ret = 0; | |
791 | int doit = task == current; | |
792 | int cpu; | |
793 | ||
794 | switch (code) { | |
795 | case ARCH_SET_GS: | |
84929801 | 796 | if (addr >= TASK_SIZE_OF(task)) |
1da177e4 LT |
797 | return -EPERM; |
798 | cpu = get_cpu(); | |
799 | /* handle small bases via the GDT because that's faster to | |
800 | switch. */ | |
801 | if (addr <= 0xffffffff) { | |
802 | set_32bit_tls(task, GS_TLS, addr); | |
803 | if (doit) { | |
804 | load_TLS(&task->thread, cpu); | |
805 | load_gs_index(GS_TLS_SEL); | |
806 | } | |
807 | task->thread.gsindex = GS_TLS_SEL; | |
808 | task->thread.gs = 0; | |
809 | } else { | |
810 | task->thread.gsindex = 0; | |
811 | task->thread.gs = addr; | |
812 | if (doit) { | |
a88cde13 AK |
813 | load_gs_index(0); |
814 | ret = checking_wrmsrl(MSR_KERNEL_GS_BASE, addr); | |
1da177e4 LT |
815 | } |
816 | } | |
817 | put_cpu(); | |
818 | break; | |
819 | case ARCH_SET_FS: | |
820 | /* Not strictly needed for fs, but do it for symmetry | |
821 | with gs */ | |
84929801 | 822 | if (addr >= TASK_SIZE_OF(task)) |
1da177e4 LT |
823 | return -EPERM; |
824 | cpu = get_cpu(); | |
825 | /* handle small bases via the GDT because that's faster to | |
826 | switch. */ | |
827 | if (addr <= 0xffffffff) { | |
828 | set_32bit_tls(task, FS_TLS, addr); | |
829 | if (doit) { | |
830 | load_TLS(&task->thread, cpu); | |
a88cde13 | 831 | asm volatile("movl %0,%%fs" :: "r"(FS_TLS_SEL)); |
1da177e4 LT |
832 | } |
833 | task->thread.fsindex = FS_TLS_SEL; | |
834 | task->thread.fs = 0; | |
835 | } else { | |
836 | task->thread.fsindex = 0; | |
837 | task->thread.fs = addr; | |
838 | if (doit) { | |
839 | /* set the selector to 0 to not confuse | |
840 | __switch_to */ | |
a88cde13 AK |
841 | asm volatile("movl %0,%%fs" :: "r" (0)); |
842 | ret = checking_wrmsrl(MSR_FS_BASE, addr); | |
1da177e4 LT |
843 | } |
844 | } | |
845 | put_cpu(); | |
846 | break; | |
847 | case ARCH_GET_FS: { | |
848 | unsigned long base; | |
849 | if (task->thread.fsindex == FS_TLS_SEL) | |
850 | base = read_32bit_tls(task, FS_TLS); | |
a88cde13 | 851 | else if (doit) |
1da177e4 | 852 | rdmsrl(MSR_FS_BASE, base); |
a88cde13 | 853 | else |
1da177e4 LT |
854 | base = task->thread.fs; |
855 | ret = put_user(base, (unsigned long __user *)addr); | |
856 | break; | |
857 | } | |
858 | case ARCH_GET_GS: { | |
859 | unsigned long base; | |
97c2803c | 860 | unsigned gsindex; |
1da177e4 LT |
861 | if (task->thread.gsindex == GS_TLS_SEL) |
862 | base = read_32bit_tls(task, GS_TLS); | |
97c2803c JB |
863 | else if (doit) { |
864 | asm("movl %%gs,%0" : "=r" (gsindex)); | |
865 | if (gsindex) | |
866 | rdmsrl(MSR_KERNEL_GS_BASE, base); | |
867 | else | |
868 | base = task->thread.gs; | |
869 | } | |
a88cde13 | 870 | else |
1da177e4 LT |
871 | base = task->thread.gs; |
872 | ret = put_user(base, (unsigned long __user *)addr); | |
873 | break; | |
874 | } | |
875 | ||
876 | default: | |
877 | ret = -EINVAL; | |
878 | break; | |
879 | } | |
880 | ||
881 | return ret; | |
882 | } | |
883 | ||
884 | long sys_arch_prctl(int code, unsigned long addr) | |
885 | { | |
886 | return do_arch_prctl(current, code, addr); | |
887 | } | |
888 | ||
889 | /* | |
890 | * Capture the user space registers if the task is not running (in user space) | |
891 | */ | |
892 | int dump_task_regs(struct task_struct *tsk, elf_gregset_t *regs) | |
893 | { | |
894 | struct pt_regs *pp, ptregs; | |
895 | ||
bb049232 | 896 | pp = task_pt_regs(tsk); |
1da177e4 LT |
897 | |
898 | ptregs = *pp; | |
899 | ptregs.cs &= 0xffff; | |
900 | ptregs.ss &= 0xffff; | |
901 | ||
902 | elf_core_copy_regs(regs, &ptregs); | |
903 | ||
904 | return 1; | |
905 | } | |
906 | ||
907 | unsigned long arch_align_stack(unsigned long sp) | |
908 | { | |
c16b63e0 | 909 | if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space) |
1da177e4 LT |
910 | sp -= get_random_int() % 8192; |
911 | return sp & ~0xf; | |
912 | } | |
c1d171a0 JK |
913 | |
914 | unsigned long arch_randomize_brk(struct mm_struct *mm) | |
915 | { | |
916 | unsigned long range_end = mm->brk + 0x02000000; | |
917 | return randomize_range(mm->brk, range_end, 0) ? : mm->brk; | |
918 | } | |
919 |