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 | ||
8 | /* | |
9 | * This file handles the architecture-dependent parts of process handling.. | |
10 | */ | |
11 | ||
12 | #include <stdarg.h> | |
13 | ||
f3705136 | 14 | #include <linux/cpu.h> |
1da177e4 LT |
15 | #include <linux/errno.h> |
16 | #include <linux/sched.h> | |
17 | #include <linux/fs.h> | |
18 | #include <linux/kernel.h> | |
19 | #include <linux/mm.h> | |
20 | #include <linux/elfcore.h> | |
21 | #include <linux/smp.h> | |
1da177e4 LT |
22 | #include <linux/stddef.h> |
23 | #include <linux/slab.h> | |
24 | #include <linux/vmalloc.h> | |
25 | #include <linux/user.h> | |
26 | #include <linux/a.out.h> | |
27 | #include <linux/interrupt.h> | |
1da177e4 LT |
28 | #include <linux/utsname.h> |
29 | #include <linux/delay.h> | |
30 | #include <linux/reboot.h> | |
31 | #include <linux/init.h> | |
32 | #include <linux/mc146818rtc.h> | |
33 | #include <linux/module.h> | |
34 | #include <linux/kallsyms.h> | |
35 | #include <linux/ptrace.h> | |
36 | #include <linux/random.h> | |
c16b63e0 | 37 | #include <linux/personality.h> |
74167347 | 38 | #include <linux/tick.h> |
7c3576d2 | 39 | #include <linux/percpu.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/ldt.h> | |
46 | #include <asm/processor.h> | |
47 | #include <asm/i387.h> | |
1da177e4 | 48 | #include <asm/desc.h> |
64ca9004 | 49 | #include <asm/vm86.h> |
1da177e4 LT |
50 | #ifdef CONFIG_MATH_EMULATION |
51 | #include <asm/math_emu.h> | |
52 | #endif | |
53 | ||
1da177e4 LT |
54 | #include <linux/err.h> |
55 | ||
f3705136 ZM |
56 | #include <asm/tlbflush.h> |
57 | #include <asm/cpu.h> | |
718fc13b | 58 | #include <asm/kdebug.h> |
f3705136 | 59 | |
1da177e4 LT |
60 | asmlinkage void ret_from_fork(void) __asm__("ret_from_fork"); |
61 | ||
62 | static int hlt_counter; | |
63 | ||
64 | unsigned long boot_option_idle_override = 0; | |
65 | EXPORT_SYMBOL(boot_option_idle_override); | |
66 | ||
7c3576d2 JF |
67 | DEFINE_PER_CPU(struct task_struct *, current_task) = &init_task; |
68 | EXPORT_PER_CPU_SYMBOL(current_task); | |
69 | ||
70 | DEFINE_PER_CPU(int, cpu_number); | |
71 | EXPORT_PER_CPU_SYMBOL(cpu_number); | |
72 | ||
1da177e4 LT |
73 | /* |
74 | * Return saved PC of a blocked thread. | |
75 | */ | |
76 | unsigned long thread_saved_pc(struct task_struct *tsk) | |
77 | { | |
faca6227 | 78 | return ((unsigned long *)tsk->thread.sp)[3]; |
1da177e4 LT |
79 | } |
80 | ||
81 | /* | |
82 | * Powermanagement idle function, if any.. | |
83 | */ | |
84 | void (*pm_idle)(void); | |
129f6946 | 85 | EXPORT_SYMBOL(pm_idle); |
1da177e4 LT |
86 | static DEFINE_PER_CPU(unsigned int, cpu_idle_state); |
87 | ||
88 | void disable_hlt(void) | |
89 | { | |
90 | hlt_counter++; | |
91 | } | |
92 | ||
93 | EXPORT_SYMBOL(disable_hlt); | |
94 | ||
95 | void enable_hlt(void) | |
96 | { | |
97 | hlt_counter--; | |
98 | } | |
99 | ||
100 | EXPORT_SYMBOL(enable_hlt); | |
101 | ||
102 | /* | |
103 | * We use this if we don't have any better | |
104 | * idle routine.. | |
105 | */ | |
106 | void default_idle(void) | |
107 | { | |
108 | if (!hlt_counter && boot_cpu_data.hlt_works_ok) { | |
495ab9c0 | 109 | current_thread_info()->status &= ~TS_POLLING; |
0888f06a IM |
110 | /* |
111 | * TS_POLLING-cleared state must be visible before we | |
112 | * test NEED_RESCHED: | |
113 | */ | |
114 | smp_mb(); | |
115 | ||
72690a21 | 116 | local_irq_disable(); |
5ee613b6 IM |
117 | if (!need_resched()) { |
118 | ktime_t t0, t1; | |
119 | u64 t0n, t1n; | |
120 | ||
121 | t0 = ktime_get(); | |
122 | t0n = ktime_to_ns(t0); | |
72690a21 | 123 | safe_halt(); /* enables interrupts racelessly */ |
5ee613b6 IM |
124 | local_irq_disable(); |
125 | t1 = ktime_get(); | |
126 | t1n = ktime_to_ns(t1); | |
127 | sched_clock_idle_wakeup_event(t1n - t0n); | |
128 | } | |
129 | local_irq_enable(); | |
495ab9c0 | 130 | current_thread_info()->status |= TS_POLLING; |
1da177e4 | 131 | } else { |
72690a21 AK |
132 | /* loop is done by the caller */ |
133 | cpu_relax(); | |
1da177e4 LT |
134 | } |
135 | } | |
129f6946 AD |
136 | #ifdef CONFIG_APM_MODULE |
137 | EXPORT_SYMBOL(default_idle); | |
138 | #endif | |
1da177e4 LT |
139 | |
140 | /* | |
141 | * On SMP it's slightly faster (but much more power-consuming!) | |
142 | * to poll the ->work.need_resched flag instead of waiting for the | |
143 | * cross-CPU IPI to arrive. Use this option with caution. | |
144 | */ | |
6612538c | 145 | static void poll_idle(void) |
1da177e4 | 146 | { |
72690a21 | 147 | cpu_relax(); |
1da177e4 LT |
148 | } |
149 | ||
f3705136 ZM |
150 | #ifdef CONFIG_HOTPLUG_CPU |
151 | #include <asm/nmi.h> | |
152 | /* We don't actually take CPU down, just spin without interrupts. */ | |
153 | static inline void play_dead(void) | |
154 | { | |
e1367daf LS |
155 | /* This must be done before dead CPU ack */ |
156 | cpu_exit_clear(); | |
157 | wbinvd(); | |
158 | mb(); | |
f3705136 ZM |
159 | /* Ack it */ |
160 | __get_cpu_var(cpu_state) = CPU_DEAD; | |
161 | ||
e1367daf LS |
162 | /* |
163 | * With physical CPU hotplug, we should halt the cpu | |
164 | */ | |
f3705136 | 165 | local_irq_disable(); |
e1367daf | 166 | while (1) |
f2ab4461 | 167 | halt(); |
f3705136 ZM |
168 | } |
169 | #else | |
170 | static inline void play_dead(void) | |
171 | { | |
172 | BUG(); | |
173 | } | |
174 | #endif /* CONFIG_HOTPLUG_CPU */ | |
175 | ||
1da177e4 LT |
176 | /* |
177 | * The idle thread. There's no useful work to be | |
178 | * done, so just try to conserve power and have a | |
179 | * low exit latency (ie sit in a loop waiting for | |
180 | * somebody to say that they'd like to reschedule) | |
181 | */ | |
f3705136 | 182 | void cpu_idle(void) |
1da177e4 | 183 | { |
5bfb5d69 | 184 | int cpu = smp_processor_id(); |
f3705136 | 185 | |
495ab9c0 | 186 | current_thread_info()->status |= TS_POLLING; |
64c7c8f8 | 187 | |
1da177e4 LT |
188 | /* endless idle loop with no priority at all */ |
189 | while (1) { | |
74167347 | 190 | tick_nohz_stop_sched_tick(); |
1da177e4 LT |
191 | while (!need_resched()) { |
192 | void (*idle)(void); | |
193 | ||
194 | if (__get_cpu_var(cpu_idle_state)) | |
195 | __get_cpu_var(cpu_idle_state) = 0; | |
196 | ||
f1d1a842 | 197 | check_pgt_cache(); |
1da177e4 LT |
198 | rmb(); |
199 | idle = pm_idle; | |
200 | ||
0723a69a BL |
201 | if (rcu_pending(cpu)) |
202 | rcu_check_callbacks(cpu, 0); | |
203 | ||
1da177e4 LT |
204 | if (!idle) |
205 | idle = default_idle; | |
206 | ||
f3705136 ZM |
207 | if (cpu_is_offline(cpu)) |
208 | play_dead(); | |
209 | ||
1da177e4 LT |
210 | __get_cpu_var(irq_stat).idle_timestamp = jiffies; |
211 | idle(); | |
212 | } | |
74167347 | 213 | tick_nohz_restart_sched_tick(); |
5bfb5d69 | 214 | preempt_enable_no_resched(); |
1da177e4 | 215 | schedule(); |
5bfb5d69 | 216 | preempt_disable(); |
1da177e4 LT |
217 | } |
218 | } | |
219 | ||
40d6a146 SR |
220 | static void do_nothing(void *unused) |
221 | { | |
222 | } | |
223 | ||
1da177e4 LT |
224 | void cpu_idle_wait(void) |
225 | { | |
226 | unsigned int cpu, this_cpu = get_cpu(); | |
dc1829a4 | 227 | cpumask_t map, tmp = current->cpus_allowed; |
1da177e4 LT |
228 | |
229 | set_cpus_allowed(current, cpumask_of_cpu(this_cpu)); | |
230 | put_cpu(); | |
231 | ||
232 | cpus_clear(map); | |
233 | for_each_online_cpu(cpu) { | |
234 | per_cpu(cpu_idle_state, cpu) = 1; | |
235 | cpu_set(cpu, map); | |
236 | } | |
237 | ||
238 | __get_cpu_var(cpu_idle_state) = 0; | |
239 | ||
240 | wmb(); | |
241 | do { | |
242 | ssleep(1); | |
243 | for_each_online_cpu(cpu) { | |
244 | if (cpu_isset(cpu, map) && !per_cpu(cpu_idle_state, cpu)) | |
245 | cpu_clear(cpu, map); | |
246 | } | |
247 | cpus_and(map, map, cpu_online_map); | |
40d6a146 SR |
248 | /* |
249 | * We waited 1 sec, if a CPU still did not call idle | |
250 | * it may be because it is in idle and not waking up | |
251 | * because it has nothing to do. | |
252 | * Give all the remaining CPUS a kick. | |
253 | */ | |
254 | smp_call_function_mask(map, do_nothing, 0, 0); | |
1da177e4 | 255 | } while (!cpus_empty(map)); |
dc1829a4 IM |
256 | |
257 | set_cpus_allowed(current, tmp); | |
1da177e4 LT |
258 | } |
259 | EXPORT_SYMBOL_GPL(cpu_idle_wait); | |
260 | ||
261 | /* | |
262 | * This uses new MONITOR/MWAIT instructions on P4 processors with PNI, | |
263 | * which can obviate IPI to trigger checking of need_resched. | |
264 | * We execute MONITOR against need_resched and enter optimized wait state | |
265 | * through MWAIT. Whenever someone changes need_resched, we would be woken | |
266 | * up from MWAIT (without an IPI). | |
991528d7 VP |
267 | * |
268 | * New with Core Duo processors, MWAIT can take some hints based on CPU | |
269 | * capability. | |
1da177e4 | 270 | */ |
65ea5b03 | 271 | void mwait_idle_with_hints(unsigned long ax, unsigned long cx) |
1da177e4 | 272 | { |
991528d7 | 273 | if (!need_resched()) { |
64c7c8f8 NP |
274 | __monitor((void *)¤t_thread_info()->flags, 0, 0); |
275 | smp_mb(); | |
991528d7 | 276 | if (!need_resched()) |
65ea5b03 | 277 | __mwait(ax, cx); |
1da177e4 LT |
278 | } |
279 | } | |
280 | ||
991528d7 VP |
281 | /* Default MONITOR/MWAIT with no hints, used for default C1 state */ |
282 | static void mwait_idle(void) | |
283 | { | |
284 | local_irq_enable(); | |
72690a21 | 285 | mwait_idle_with_hints(0, 0); |
991528d7 VP |
286 | } |
287 | ||
0c07ee38 AK |
288 | static int mwait_usable(const struct cpuinfo_x86 *c) |
289 | { | |
290 | if (force_mwait) | |
291 | return 1; | |
292 | /* Any C1 states supported? */ | |
293 | return c->cpuid_level >= 5 && ((cpuid_edx(5) >> 4) & 0xf) > 0; | |
294 | } | |
295 | ||
3446fa05 | 296 | void __cpuinit select_idle_routine(const struct cpuinfo_x86 *c) |
1da177e4 | 297 | { |
0c07ee38 | 298 | if (cpu_has(c, X86_FEATURE_MWAIT) && mwait_usable(c)) { |
1da177e4 LT |
299 | printk("monitor/mwait feature present.\n"); |
300 | /* | |
301 | * Skip, if setup has overridden idle. | |
302 | * One CPU supports mwait => All CPUs supports mwait | |
303 | */ | |
304 | if (!pm_idle) { | |
305 | printk("using mwait in idle threads.\n"); | |
306 | pm_idle = mwait_idle; | |
307 | } | |
308 | } | |
309 | } | |
310 | ||
f039b754 | 311 | static int __init idle_setup(char *str) |
1da177e4 | 312 | { |
f039b754 | 313 | if (!strcmp(str, "poll")) { |
1da177e4 LT |
314 | printk("using polling idle threads.\n"); |
315 | pm_idle = poll_idle; | |
316 | #ifdef CONFIG_X86_SMP | |
317 | if (smp_num_siblings > 1) | |
318 | printk("WARNING: polling idle and HT enabled, performance may degrade.\n"); | |
319 | #endif | |
f039b754 AK |
320 | } else if (!strcmp(str, "mwait")) |
321 | force_mwait = 1; | |
322 | else | |
323 | return -1; | |
1da177e4 LT |
324 | |
325 | boot_option_idle_override = 1; | |
f039b754 | 326 | return 0; |
1da177e4 | 327 | } |
f039b754 | 328 | early_param("idle", idle_setup); |
1da177e4 | 329 | |
9d975ebd | 330 | void __show_registers(struct pt_regs *regs, int all) |
1da177e4 LT |
331 | { |
332 | unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L; | |
bb1995d5 | 333 | unsigned long d0, d1, d2, d3, d6, d7; |
65ea5b03 | 334 | unsigned long sp; |
9d975ebd PE |
335 | unsigned short ss, gs; |
336 | ||
337 | if (user_mode_vm(regs)) { | |
65ea5b03 PA |
338 | sp = regs->sp; |
339 | ss = regs->ss & 0xffff; | |
9d975ebd PE |
340 | savesegment(gs, gs); |
341 | } else { | |
65ea5b03 | 342 | sp = (unsigned long) (®s->sp); |
9d975ebd PE |
343 | savesegment(ss, ss); |
344 | savesegment(gs, gs); | |
345 | } | |
1da177e4 LT |
346 | |
347 | printk("\n"); | |
60812a4a LT |
348 | printk("Pid: %d, comm: %s %s (%s %.*s)\n", |
349 | task_pid_nr(current), current->comm, | |
9d975ebd PE |
350 | print_tainted(), init_utsname()->release, |
351 | (int)strcspn(init_utsname()->version, " "), | |
352 | init_utsname()->version); | |
353 | ||
354 | printk("EIP: %04x:[<%08lx>] EFLAGS: %08lx CPU: %d\n", | |
65ea5b03 | 355 | 0xffff & regs->cs, regs->ip, regs->flags, |
9d975ebd | 356 | smp_processor_id()); |
65ea5b03 | 357 | print_symbol("EIP is at %s\n", regs->ip); |
1da177e4 | 358 | |
1da177e4 | 359 | printk("EAX: %08lx EBX: %08lx ECX: %08lx EDX: %08lx\n", |
65ea5b03 | 360 | regs->ax, regs->bx, regs->cx, regs->dx); |
9d975ebd | 361 | printk("ESI: %08lx EDI: %08lx EBP: %08lx ESP: %08lx\n", |
65ea5b03 | 362 | regs->si, regs->di, regs->bp, sp); |
9d975ebd | 363 | printk(" DS: %04x ES: %04x FS: %04x GS: %04x SS: %04x\n", |
65ea5b03 PA |
364 | regs->ds & 0xffff, regs->es & 0xffff, |
365 | regs->fs & 0xffff, gs, ss); | |
9d975ebd PE |
366 | |
367 | if (!all) | |
368 | return; | |
1da177e4 | 369 | |
4bb0d3ec ZA |
370 | cr0 = read_cr0(); |
371 | cr2 = read_cr2(); | |
372 | cr3 = read_cr3(); | |
ff6e8c0d | 373 | cr4 = read_cr4_safe(); |
9d975ebd PE |
374 | printk("CR0: %08lx CR2: %08lx CR3: %08lx CR4: %08lx\n", |
375 | cr0, cr2, cr3, cr4); | |
bb1995d5 AS |
376 | |
377 | get_debugreg(d0, 0); | |
378 | get_debugreg(d1, 1); | |
379 | get_debugreg(d2, 2); | |
380 | get_debugreg(d3, 3); | |
381 | printk("DR0: %08lx DR1: %08lx DR2: %08lx DR3: %08lx\n", | |
382 | d0, d1, d2, d3); | |
9d975ebd | 383 | |
bb1995d5 AS |
384 | get_debugreg(d6, 6); |
385 | get_debugreg(d7, 7); | |
9d975ebd PE |
386 | printk("DR6: %08lx DR7: %08lx\n", |
387 | d6, d7); | |
388 | } | |
bb1995d5 | 389 | |
9d975ebd PE |
390 | void show_regs(struct pt_regs *regs) |
391 | { | |
392 | __show_registers(regs, 1); | |
5bc27dc2 | 393 | show_trace(NULL, regs, ®s->sp, regs->bp); |
1da177e4 LT |
394 | } |
395 | ||
396 | /* | |
65ea5b03 PA |
397 | * This gets run with %bx containing the |
398 | * function to call, and %dx containing | |
1da177e4 LT |
399 | * the "args". |
400 | */ | |
401 | extern void kernel_thread_helper(void); | |
1da177e4 LT |
402 | |
403 | /* | |
404 | * Create a kernel thread | |
405 | */ | |
406 | int kernel_thread(int (*fn)(void *), void * arg, unsigned long flags) | |
407 | { | |
408 | struct pt_regs regs; | |
409 | ||
410 | memset(®s, 0, sizeof(regs)); | |
411 | ||
65ea5b03 PA |
412 | regs.bx = (unsigned long) fn; |
413 | regs.dx = (unsigned long) arg; | |
1da177e4 | 414 | |
65ea5b03 PA |
415 | regs.ds = __USER_DS; |
416 | regs.es = __USER_DS; | |
417 | regs.fs = __KERNEL_PERCPU; | |
418 | regs.orig_ax = -1; | |
419 | regs.ip = (unsigned long) kernel_thread_helper; | |
420 | regs.cs = __KERNEL_CS | get_kernel_rpl(); | |
421 | regs.flags = X86_EFLAGS_IF | X86_EFLAGS_SF | X86_EFLAGS_PF | 0x2; | |
1da177e4 LT |
422 | |
423 | /* Ok, create the new process.. */ | |
8cf2c519 | 424 | return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, ®s, 0, NULL, NULL); |
1da177e4 | 425 | } |
129f6946 | 426 | EXPORT_SYMBOL(kernel_thread); |
1da177e4 LT |
427 | |
428 | /* | |
429 | * Free current thread data structures etc.. | |
430 | */ | |
431 | void exit_thread(void) | |
432 | { | |
1da177e4 | 433 | /* The process may have allocated an io port bitmap... nuke it. */ |
b3cf2576 SE |
434 | if (unlikely(test_thread_flag(TIF_IO_BITMAP))) { |
435 | struct task_struct *tsk = current; | |
436 | struct thread_struct *t = &tsk->thread; | |
1da177e4 LT |
437 | int cpu = get_cpu(); |
438 | struct tss_struct *tss = &per_cpu(init_tss, cpu); | |
439 | ||
440 | kfree(t->io_bitmap_ptr); | |
441 | t->io_bitmap_ptr = NULL; | |
b3cf2576 | 442 | clear_thread_flag(TIF_IO_BITMAP); |
1da177e4 LT |
443 | /* |
444 | * Careful, clear this in the TSS too: | |
445 | */ | |
446 | memset(tss->io_bitmap, 0xff, tss->io_bitmap_max); | |
447 | t->io_bitmap_max = 0; | |
448 | tss->io_bitmap_owner = NULL; | |
449 | tss->io_bitmap_max = 0; | |
a75c54f9 | 450 | tss->x86_tss.io_bitmap_base = INVALID_IO_BITMAP_OFFSET; |
1da177e4 LT |
451 | put_cpu(); |
452 | } | |
453 | } | |
454 | ||
455 | void flush_thread(void) | |
456 | { | |
457 | struct task_struct *tsk = current; | |
458 | ||
0f534093 RM |
459 | tsk->thread.debugreg0 = 0; |
460 | tsk->thread.debugreg1 = 0; | |
461 | tsk->thread.debugreg2 = 0; | |
462 | tsk->thread.debugreg3 = 0; | |
463 | tsk->thread.debugreg6 = 0; | |
464 | tsk->thread.debugreg7 = 0; | |
1da177e4 | 465 | memset(tsk->thread.tls_array, 0, sizeof(tsk->thread.tls_array)); |
b3cf2576 | 466 | clear_tsk_thread_flag(tsk, TIF_DEBUG); |
1da177e4 LT |
467 | /* |
468 | * Forget coprocessor state.. | |
469 | */ | |
470 | clear_fpu(tsk); | |
471 | clear_used_math(); | |
472 | } | |
473 | ||
474 | void release_thread(struct task_struct *dead_task) | |
475 | { | |
2684927c | 476 | BUG_ON(dead_task->mm); |
1da177e4 LT |
477 | release_vm86_irqs(dead_task); |
478 | } | |
479 | ||
480 | /* | |
481 | * This gets called before we allocate a new thread and copy | |
482 | * the current task into it. | |
483 | */ | |
484 | void prepare_to_copy(struct task_struct *tsk) | |
485 | { | |
486 | unlazy_fpu(tsk); | |
487 | } | |
488 | ||
65ea5b03 | 489 | int copy_thread(int nr, unsigned long clone_flags, unsigned long sp, |
1da177e4 LT |
490 | unsigned long unused, |
491 | struct task_struct * p, struct pt_regs * regs) | |
492 | { | |
493 | struct pt_regs * childregs; | |
494 | struct task_struct *tsk; | |
495 | int err; | |
496 | ||
07b047fc | 497 | childregs = task_pt_regs(p); |
f48d9663 | 498 | *childregs = *regs; |
65ea5b03 PA |
499 | childregs->ax = 0; |
500 | childregs->sp = sp; | |
f48d9663 | 501 | |
faca6227 PA |
502 | p->thread.sp = (unsigned long) childregs; |
503 | p->thread.sp0 = (unsigned long) (childregs+1); | |
1da177e4 | 504 | |
faca6227 | 505 | p->thread.ip = (unsigned long) ret_from_fork; |
1da177e4 | 506 | |
6612538c | 507 | savesegment(gs, p->thread.gs); |
1da177e4 LT |
508 | |
509 | tsk = current; | |
b3cf2576 | 510 | if (unlikely(test_tsk_thread_flag(tsk, TIF_IO_BITMAP))) { |
52978be6 AD |
511 | p->thread.io_bitmap_ptr = kmemdup(tsk->thread.io_bitmap_ptr, |
512 | IO_BITMAP_BYTES, GFP_KERNEL); | |
1da177e4 LT |
513 | if (!p->thread.io_bitmap_ptr) { |
514 | p->thread.io_bitmap_max = 0; | |
515 | return -ENOMEM; | |
516 | } | |
b3cf2576 | 517 | set_tsk_thread_flag(p, TIF_IO_BITMAP); |
1da177e4 LT |
518 | } |
519 | ||
efd1ca52 RM |
520 | err = 0; |
521 | ||
1da177e4 LT |
522 | /* |
523 | * Set a new TLS for the child thread? | |
524 | */ | |
efd1ca52 RM |
525 | if (clone_flags & CLONE_SETTLS) |
526 | err = do_set_thread_area(p, -1, | |
65ea5b03 | 527 | (struct user_desc __user *)childregs->si, 0); |
1da177e4 | 528 | |
1da177e4 LT |
529 | if (err && p->thread.io_bitmap_ptr) { |
530 | kfree(p->thread.io_bitmap_ptr); | |
531 | p->thread.io_bitmap_max = 0; | |
532 | } | |
533 | return err; | |
534 | } | |
535 | ||
536 | /* | |
537 | * fill in the user structure for a core dump.. | |
538 | */ | |
539 | void dump_thread(struct pt_regs * regs, struct user * dump) | |
540 | { | |
153d5f2e | 541 | u16 gs; |
1da177e4 LT |
542 | |
543 | /* changed the size calculations - should hopefully work better. lbt */ | |
544 | dump->magic = CMAGIC; | |
545 | dump->start_code = 0; | |
65ea5b03 | 546 | dump->start_stack = regs->sp & ~(PAGE_SIZE - 1); |
1da177e4 LT |
547 | dump->u_tsize = ((unsigned long) current->mm->end_code) >> PAGE_SHIFT; |
548 | dump->u_dsize = ((unsigned long) (current->mm->brk + (PAGE_SIZE-1))) >> PAGE_SHIFT; | |
549 | dump->u_dsize -= dump->u_tsize; | |
550 | dump->u_ssize = 0; | |
0f534093 RM |
551 | dump->u_debugreg[0] = current->thread.debugreg0; |
552 | dump->u_debugreg[1] = current->thread.debugreg1; | |
553 | dump->u_debugreg[2] = current->thread.debugreg2; | |
554 | dump->u_debugreg[3] = current->thread.debugreg3; | |
555 | dump->u_debugreg[4] = 0; | |
556 | dump->u_debugreg[5] = 0; | |
557 | dump->u_debugreg[6] = current->thread.debugreg6; | |
558 | dump->u_debugreg[7] = current->thread.debugreg7; | |
1da177e4 LT |
559 | |
560 | if (dump->start_stack < TASK_SIZE) | |
561 | dump->u_ssize = ((unsigned long) (TASK_SIZE - dump->start_stack)) >> PAGE_SHIFT; | |
562 | ||
153d5f2e PA |
563 | dump->regs.bx = regs->bx; |
564 | dump->regs.cx = regs->cx; | |
565 | dump->regs.dx = regs->dx; | |
566 | dump->regs.si = regs->si; | |
567 | dump->regs.di = regs->di; | |
568 | dump->regs.bp = regs->bp; | |
569 | dump->regs.ax = regs->ax; | |
570 | dump->regs.ds = (u16)regs->ds; | |
571 | dump->regs.es = (u16)regs->es; | |
572 | dump->regs.fs = (u16)regs->fs; | |
573 | savesegment(gs,gs); | |
574 | dump->regs.orig_ax = regs->orig_ax; | |
575 | dump->regs.ip = regs->ip; | |
576 | dump->regs.cs = (u16)regs->cs; | |
577 | dump->regs.flags = regs->flags; | |
578 | dump->regs.sp = regs->sp; | |
579 | dump->regs.ss = (u16)regs->ss; | |
1da177e4 LT |
580 | |
581 | dump->u_fpvalid = dump_fpu (regs, &dump->i387); | |
582 | } | |
129f6946 | 583 | EXPORT_SYMBOL(dump_thread); |
1da177e4 | 584 | |
cf99abac | 585 | #ifdef CONFIG_SECCOMP |
bdb4f156 | 586 | static void hard_disable_TSC(void) |
cf99abac AA |
587 | { |
588 | write_cr4(read_cr4() | X86_CR4_TSD); | |
589 | } | |
590 | void disable_TSC(void) | |
591 | { | |
592 | preempt_disable(); | |
593 | if (!test_and_set_thread_flag(TIF_NOTSC)) | |
594 | /* | |
595 | * Must flip the CPU state synchronously with | |
596 | * TIF_NOTSC in the current running context. | |
597 | */ | |
598 | hard_disable_TSC(); | |
599 | preempt_enable(); | |
600 | } | |
bdb4f156 | 601 | static void hard_enable_TSC(void) |
cf99abac AA |
602 | { |
603 | write_cr4(read_cr4() & ~X86_CR4_TSD); | |
604 | } | |
605 | #endif /* CONFIG_SECCOMP */ | |
606 | ||
607 | static noinline void | |
608 | __switch_to_xtra(struct task_struct *prev_p, struct task_struct *next_p, | |
609 | struct tss_struct *tss) | |
1da177e4 | 610 | { |
7e991604 | 611 | struct thread_struct *prev, *next; |
eee3af4a | 612 | unsigned long debugctl; |
b3cf2576 | 613 | |
7e991604 | 614 | prev = &prev_p->thread; |
b3cf2576 SE |
615 | next = &next_p->thread; |
616 | ||
eee3af4a MM |
617 | debugctl = prev->debugctlmsr; |
618 | if (next->ds_area_msr != prev->ds_area_msr) { | |
619 | /* we clear debugctl to make sure DS | |
620 | * is not in use when we change it */ | |
621 | debugctl = 0; | |
622 | wrmsrl(MSR_IA32_DEBUGCTLMSR, 0); | |
623 | wrmsr(MSR_IA32_DS_AREA, next->ds_area_msr, 0); | |
624 | } | |
625 | ||
626 | if (next->debugctlmsr != debugctl) | |
7e991604 RM |
627 | wrmsr(MSR_IA32_DEBUGCTLMSR, next->debugctlmsr, 0); |
628 | ||
b3cf2576 | 629 | if (test_tsk_thread_flag(next_p, TIF_DEBUG)) { |
0f534093 RM |
630 | set_debugreg(next->debugreg0, 0); |
631 | set_debugreg(next->debugreg1, 1); | |
632 | set_debugreg(next->debugreg2, 2); | |
633 | set_debugreg(next->debugreg3, 3); | |
b3cf2576 | 634 | /* no 4 and 5 */ |
0f534093 RM |
635 | set_debugreg(next->debugreg6, 6); |
636 | set_debugreg(next->debugreg7, 7); | |
b3cf2576 SE |
637 | } |
638 | ||
cf99abac AA |
639 | #ifdef CONFIG_SECCOMP |
640 | if (test_tsk_thread_flag(prev_p, TIF_NOTSC) ^ | |
641 | test_tsk_thread_flag(next_p, TIF_NOTSC)) { | |
642 | /* prev and next are different */ | |
643 | if (test_tsk_thread_flag(next_p, TIF_NOTSC)) | |
644 | hard_disable_TSC(); | |
645 | else | |
646 | hard_enable_TSC(); | |
647 | } | |
648 | #endif | |
649 | ||
eee3af4a MM |
650 | if (test_tsk_thread_flag(prev_p, TIF_BTS_TRACE_TS)) |
651 | ptrace_bts_take_timestamp(prev_p, BTS_TASK_DEPARTS); | |
652 | ||
653 | if (test_tsk_thread_flag(next_p, TIF_BTS_TRACE_TS)) | |
654 | ptrace_bts_take_timestamp(next_p, BTS_TASK_ARRIVES); | |
655 | ||
656 | ||
b3cf2576 | 657 | if (!test_tsk_thread_flag(next_p, TIF_IO_BITMAP)) { |
1da177e4 LT |
658 | /* |
659 | * Disable the bitmap via an invalid offset. We still cache | |
660 | * the previous bitmap owner and the IO bitmap contents: | |
661 | */ | |
a75c54f9 | 662 | tss->x86_tss.io_bitmap_base = INVALID_IO_BITMAP_OFFSET; |
1da177e4 LT |
663 | return; |
664 | } | |
b3cf2576 | 665 | |
1da177e4 LT |
666 | if (likely(next == tss->io_bitmap_owner)) { |
667 | /* | |
668 | * Previous owner of the bitmap (hence the bitmap content) | |
669 | * matches the next task, we dont have to do anything but | |
670 | * to set a valid offset in the TSS: | |
671 | */ | |
a75c54f9 | 672 | tss->x86_tss.io_bitmap_base = IO_BITMAP_OFFSET; |
1da177e4 LT |
673 | return; |
674 | } | |
675 | /* | |
676 | * Lazy TSS's I/O bitmap copy. We set an invalid offset here | |
677 | * and we let the task to get a GPF in case an I/O instruction | |
678 | * is performed. The handler of the GPF will verify that the | |
679 | * faulting task has a valid I/O bitmap and, it true, does the | |
680 | * real copy and restart the instruction. This will save us | |
681 | * redundant copies when the currently switched task does not | |
682 | * perform any I/O during its timeslice. | |
683 | */ | |
a75c54f9 | 684 | tss->x86_tss.io_bitmap_base = INVALID_IO_BITMAP_OFFSET_LAZY; |
1da177e4 | 685 | } |
1da177e4 LT |
686 | |
687 | /* | |
688 | * switch_to(x,yn) should switch tasks from x to y. | |
689 | * | |
690 | * We fsave/fwait so that an exception goes off at the right time | |
691 | * (as a call from the fsave or fwait in effect) rather than to | |
692 | * the wrong process. Lazy FP saving no longer makes any sense | |
693 | * with modern CPU's, and this simplifies a lot of things (SMP | |
694 | * and UP become the same). | |
695 | * | |
696 | * NOTE! We used to use the x86 hardware context switching. The | |
697 | * reason for not using it any more becomes apparent when you | |
698 | * try to recover gracefully from saved state that is no longer | |
699 | * valid (stale segment register values in particular). With the | |
700 | * hardware task-switch, there is no way to fix up bad state in | |
701 | * a reasonable manner. | |
702 | * | |
703 | * The fact that Intel documents the hardware task-switching to | |
704 | * be slow is a fairly red herring - this code is not noticeably | |
705 | * faster. However, there _is_ some room for improvement here, | |
706 | * so the performance issues may eventually be a valid point. | |
707 | * More important, however, is the fact that this allows us much | |
708 | * more flexibility. | |
709 | * | |
65ea5b03 | 710 | * The return value (in %ax) will be the "prev" task after |
1da177e4 LT |
711 | * the task-switch, and shows up in ret_from_fork in entry.S, |
712 | * for example. | |
713 | */ | |
75604d7f | 714 | struct task_struct * __switch_to(struct task_struct *prev_p, struct task_struct *next_p) |
1da177e4 LT |
715 | { |
716 | struct thread_struct *prev = &prev_p->thread, | |
717 | *next = &next_p->thread; | |
718 | int cpu = smp_processor_id(); | |
719 | struct tss_struct *tss = &per_cpu(init_tss, cpu); | |
720 | ||
721 | /* never put a printk in __switch_to... printk() calls wake_up*() indirectly */ | |
722 | ||
723 | __unlazy_fpu(prev_p); | |
724 | ||
acc20761 CE |
725 | |
726 | /* we're going to use this soon, after a few expensive things */ | |
727 | if (next_p->fpu_counter > 5) | |
728 | prefetch(&next->i387.fxsave); | |
729 | ||
1da177e4 | 730 | /* |
e7a2ff59 | 731 | * Reload esp0. |
1da177e4 | 732 | */ |
faca6227 | 733 | load_sp0(tss, next); |
1da177e4 LT |
734 | |
735 | /* | |
464d1a78 | 736 | * Save away %gs. No need to save %fs, as it was saved on the |
f95d47ca JF |
737 | * stack on entry. No need to save %es and %ds, as those are |
738 | * always kernel segments while inside the kernel. Doing this | |
739 | * before setting the new TLS descriptors avoids the situation | |
740 | * where we temporarily have non-reloadable segments in %fs | |
741 | * and %gs. This could be an issue if the NMI handler ever | |
742 | * used %fs or %gs (it does not today), or if the kernel is | |
743 | * running inside of a hypervisor layer. | |
1da177e4 | 744 | */ |
464d1a78 | 745 | savesegment(gs, prev->gs); |
1da177e4 LT |
746 | |
747 | /* | |
e7a2ff59 | 748 | * Load the per-thread Thread-Local Storage descriptor. |
1da177e4 | 749 | */ |
e7a2ff59 | 750 | load_TLS(next, cpu); |
1da177e4 | 751 | |
8b151144 ZA |
752 | /* |
753 | * Restore IOPL if needed. In normal use, the flags restore | |
754 | * in the switch assembly will handle this. But if the kernel | |
755 | * is running virtualized at a non-zero CPL, the popf will | |
756 | * not restore flags, so it must be done in a separate step. | |
757 | */ | |
758 | if (get_kernel_rpl() && unlikely(prev->iopl != next->iopl)) | |
759 | set_iopl_mask(next->iopl); | |
760 | ||
1da177e4 | 761 | /* |
b3cf2576 | 762 | * Now maybe handle debug registers and/or IO bitmaps |
1da177e4 | 763 | */ |
cf99abac AA |
764 | if (unlikely(task_thread_info(prev_p)->flags & _TIF_WORK_CTXSW_PREV || |
765 | task_thread_info(next_p)->flags & _TIF_WORK_CTXSW_NEXT)) | |
766 | __switch_to_xtra(prev_p, next_p, tss); | |
ffaa8bd6 | 767 | |
9226d125 ZA |
768 | /* |
769 | * Leave lazy mode, flushing any hypercalls made here. | |
770 | * This must be done before restoring TLS segments so | |
771 | * the GDT and LDT are properly updated, and must be | |
772 | * done before math_state_restore, so the TS bit is up | |
773 | * to date. | |
774 | */ | |
775 | arch_leave_lazy_cpu_mode(); | |
776 | ||
acc20761 CE |
777 | /* If the task has used fpu the last 5 timeslices, just do a full |
778 | * restore of the math state immediately to avoid the trap; the | |
779 | * chances of needing FPU soon are obviously high now | |
780 | */ | |
781 | if (next_p->fpu_counter > 5) | |
782 | math_state_restore(); | |
783 | ||
9226d125 ZA |
784 | /* |
785 | * Restore %gs if needed (which is common) | |
786 | */ | |
787 | if (prev->gs | next->gs) | |
788 | loadsegment(gs, next->gs); | |
789 | ||
7c3576d2 | 790 | x86_write_percpu(current_task, next_p); |
9226d125 | 791 | |
1da177e4 LT |
792 | return prev_p; |
793 | } | |
794 | ||
795 | asmlinkage int sys_fork(struct pt_regs regs) | |
796 | { | |
65ea5b03 | 797 | return do_fork(SIGCHLD, regs.sp, ®s, 0, NULL, NULL); |
1da177e4 LT |
798 | } |
799 | ||
800 | asmlinkage int sys_clone(struct pt_regs regs) | |
801 | { | |
802 | unsigned long clone_flags; | |
803 | unsigned long newsp; | |
804 | int __user *parent_tidptr, *child_tidptr; | |
805 | ||
65ea5b03 PA |
806 | clone_flags = regs.bx; |
807 | newsp = regs.cx; | |
808 | parent_tidptr = (int __user *)regs.dx; | |
809 | child_tidptr = (int __user *)regs.di; | |
1da177e4 | 810 | if (!newsp) |
65ea5b03 | 811 | newsp = regs.sp; |
1da177e4 LT |
812 | return do_fork(clone_flags, newsp, ®s, 0, parent_tidptr, child_tidptr); |
813 | } | |
814 | ||
815 | /* | |
816 | * This is trivial, and on the face of it looks like it | |
817 | * could equally well be done in user mode. | |
818 | * | |
819 | * Not so, for quite unobvious reasons - register pressure. | |
820 | * In user mode vfork() cannot have a stack frame, and if | |
821 | * done by calling the "clone()" system call directly, you | |
822 | * do not have enough call-clobbered registers to hold all | |
823 | * the information you need. | |
824 | */ | |
825 | asmlinkage int sys_vfork(struct pt_regs regs) | |
826 | { | |
65ea5b03 | 827 | return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, regs.sp, ®s, 0, NULL, NULL); |
1da177e4 LT |
828 | } |
829 | ||
830 | /* | |
831 | * sys_execve() executes a new program. | |
832 | */ | |
833 | asmlinkage int sys_execve(struct pt_regs regs) | |
834 | { | |
835 | int error; | |
836 | char * filename; | |
837 | ||
65ea5b03 | 838 | filename = getname((char __user *) regs.bx); |
1da177e4 LT |
839 | error = PTR_ERR(filename); |
840 | if (IS_ERR(filename)) | |
841 | goto out; | |
842 | error = do_execve(filename, | |
65ea5b03 PA |
843 | (char __user * __user *) regs.cx, |
844 | (char __user * __user *) regs.dx, | |
1da177e4 LT |
845 | ®s); |
846 | if (error == 0) { | |
1da177e4 LT |
847 | /* Make sure we don't return using sysenter.. */ |
848 | set_thread_flag(TIF_IRET); | |
849 | } | |
850 | putname(filename); | |
851 | out: | |
852 | return error; | |
853 | } | |
854 | ||
855 | #define top_esp (THREAD_SIZE - sizeof(unsigned long)) | |
856 | #define top_ebp (THREAD_SIZE - 2*sizeof(unsigned long)) | |
857 | ||
858 | unsigned long get_wchan(struct task_struct *p) | |
859 | { | |
65ea5b03 | 860 | unsigned long bp, sp, ip; |
1da177e4 LT |
861 | unsigned long stack_page; |
862 | int count = 0; | |
863 | if (!p || p == current || p->state == TASK_RUNNING) | |
864 | return 0; | |
65e0fdff | 865 | stack_page = (unsigned long)task_stack_page(p); |
faca6227 | 866 | sp = p->thread.sp; |
65ea5b03 | 867 | if (!stack_page || sp < stack_page || sp > top_esp+stack_page) |
1da177e4 | 868 | return 0; |
65ea5b03 PA |
869 | /* include/asm-i386/system.h:switch_to() pushes bp last. */ |
870 | bp = *(unsigned long *) sp; | |
1da177e4 | 871 | do { |
65ea5b03 | 872 | if (bp < stack_page || bp > top_ebp+stack_page) |
1da177e4 | 873 | return 0; |
65ea5b03 PA |
874 | ip = *(unsigned long *) (bp+4); |
875 | if (!in_sched_functions(ip)) | |
876 | return ip; | |
877 | bp = *(unsigned long *) bp; | |
1da177e4 LT |
878 | } while (count++ < 16); |
879 | return 0; | |
880 | } | |
881 | ||
1da177e4 LT |
882 | unsigned long arch_align_stack(unsigned long sp) |
883 | { | |
c16b63e0 | 884 | if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space) |
1da177e4 LT |
885 | sp -= get_random_int() % 8192; |
886 | return sp & ~0xf; | |
887 | } | |
c1d171a0 JK |
888 | |
889 | unsigned long arch_randomize_brk(struct mm_struct *mm) | |
890 | { | |
891 | unsigned long range_end = mm->brk + 0x02000000; | |
892 | return randomize_range(mm->brk, range_end, 0) ? : mm->brk; | |
893 | } |