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> | |
1da177e4 | 26 | #include <linux/interrupt.h> |
1da177e4 LT |
27 | #include <linux/utsname.h> |
28 | #include <linux/delay.h> | |
29 | #include <linux/reboot.h> | |
30 | #include <linux/init.h> | |
31 | #include <linux/mc146818rtc.h> | |
32 | #include <linux/module.h> | |
33 | #include <linux/kallsyms.h> | |
34 | #include <linux/ptrace.h> | |
35 | #include <linux/random.h> | |
c16b63e0 | 36 | #include <linux/personality.h> |
74167347 | 37 | #include <linux/tick.h> |
7c3576d2 | 38 | #include <linux/percpu.h> |
529e25f6 | 39 | #include <linux/prctl.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 LT |
48 | #include <asm/desc.h> |
49 | #ifdef CONFIG_MATH_EMULATION | |
50 | #include <asm/math_emu.h> | |
51 | #endif | |
52 | ||
1da177e4 LT |
53 | #include <linux/err.h> |
54 | ||
f3705136 ZM |
55 | #include <asm/tlbflush.h> |
56 | #include <asm/cpu.h> | |
718fc13b | 57 | #include <asm/kdebug.h> |
1eda8149 | 58 | #include <asm/idle.h> |
f3705136 | 59 | |
1da177e4 LT |
60 | asmlinkage void ret_from_fork(void) __asm__("ret_from_fork"); |
61 | ||
7c3576d2 JF |
62 | DEFINE_PER_CPU(struct task_struct *, current_task) = &init_task; |
63 | EXPORT_PER_CPU_SYMBOL(current_task); | |
64 | ||
65 | DEFINE_PER_CPU(int, cpu_number); | |
66 | EXPORT_PER_CPU_SYMBOL(cpu_number); | |
67 | ||
1da177e4 LT |
68 | /* |
69 | * Return saved PC of a blocked thread. | |
70 | */ | |
71 | unsigned long thread_saved_pc(struct task_struct *tsk) | |
72 | { | |
faca6227 | 73 | return ((unsigned long *)tsk->thread.sp)[3]; |
1da177e4 LT |
74 | } |
75 | ||
f3705136 ZM |
76 | #ifdef CONFIG_HOTPLUG_CPU |
77 | #include <asm/nmi.h> | |
1da177e4 | 78 | |
1481a3dd | 79 | static void cpu_exit_clear(void) |
1da177e4 | 80 | { |
1481a3dd | 81 | int cpu = raw_smp_processor_id(); |
1da177e4 | 82 | |
1481a3dd | 83 | idle_task_exit(); |
1da177e4 | 84 | |
1481a3dd GC |
85 | cpu_uninit(); |
86 | irq_ctx_exit(cpu); | |
1da177e4 | 87 | |
1481a3dd GC |
88 | cpu_clear(cpu, cpu_callout_map); |
89 | cpu_clear(cpu, cpu_callin_map); | |
1da177e4 | 90 | |
1481a3dd | 91 | numa_remove_cpu(cpu); |
4faac97d | 92 | c1e_remove_cpu(cpu); |
1da177e4 LT |
93 | } |
94 | ||
f3705136 ZM |
95 | /* We don't actually take CPU down, just spin without interrupts. */ |
96 | static inline void play_dead(void) | |
97 | { | |
e1367daf LS |
98 | /* This must be done before dead CPU ack */ |
99 | cpu_exit_clear(); | |
e1367daf | 100 | mb(); |
f3705136 ZM |
101 | /* Ack it */ |
102 | __get_cpu_var(cpu_state) = CPU_DEAD; | |
103 | ||
e1367daf LS |
104 | /* |
105 | * With physical CPU hotplug, we should halt the cpu | |
106 | */ | |
f3705136 | 107 | local_irq_disable(); |
394a1505 ML |
108 | /* mask all interrupts, flush any and all caches, and halt */ |
109 | wbinvd_halt(); | |
f3705136 ZM |
110 | } |
111 | #else | |
112 | static inline void play_dead(void) | |
113 | { | |
114 | BUG(); | |
115 | } | |
116 | #endif /* CONFIG_HOTPLUG_CPU */ | |
117 | ||
1da177e4 LT |
118 | /* |
119 | * The idle thread. There's no useful work to be | |
120 | * done, so just try to conserve power and have a | |
121 | * low exit latency (ie sit in a loop waiting for | |
122 | * somebody to say that they'd like to reschedule) | |
123 | */ | |
f3705136 | 124 | void cpu_idle(void) |
1da177e4 | 125 | { |
5bfb5d69 | 126 | int cpu = smp_processor_id(); |
f3705136 | 127 | |
495ab9c0 | 128 | current_thread_info()->status |= TS_POLLING; |
64c7c8f8 | 129 | |
1da177e4 LT |
130 | /* endless idle loop with no priority at all */ |
131 | while (1) { | |
b8f8c3cf | 132 | tick_nohz_stop_sched_tick(1); |
1da177e4 | 133 | while (!need_resched()) { |
1da177e4 | 134 | |
f1d1a842 | 135 | check_pgt_cache(); |
1da177e4 | 136 | rmb(); |
1da177e4 | 137 | |
0723a69a BL |
138 | if (rcu_pending(cpu)) |
139 | rcu_check_callbacks(cpu, 0); | |
140 | ||
f3705136 ZM |
141 | if (cpu_is_offline(cpu)) |
142 | play_dead(); | |
143 | ||
7f424a8b | 144 | local_irq_disable(); |
1da177e4 | 145 | __get_cpu_var(irq_stat).idle_timestamp = jiffies; |
6cd8a4bb SR |
146 | /* Don't trace irqs off for idle */ |
147 | stop_critical_timings(); | |
6ddd2a27 | 148 | pm_idle(); |
6cd8a4bb | 149 | start_critical_timings(); |
1da177e4 | 150 | } |
74167347 | 151 | tick_nohz_restart_sched_tick(); |
5bfb5d69 | 152 | preempt_enable_no_resched(); |
1da177e4 | 153 | schedule(); |
5bfb5d69 | 154 | preempt_disable(); |
1da177e4 LT |
155 | } |
156 | } | |
157 | ||
9d975ebd | 158 | void __show_registers(struct pt_regs *regs, int all) |
1da177e4 LT |
159 | { |
160 | unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L; | |
bb1995d5 | 161 | unsigned long d0, d1, d2, d3, d6, d7; |
65ea5b03 | 162 | unsigned long sp; |
9d975ebd PE |
163 | unsigned short ss, gs; |
164 | ||
165 | if (user_mode_vm(regs)) { | |
65ea5b03 PA |
166 | sp = regs->sp; |
167 | ss = regs->ss & 0xffff; | |
9d975ebd PE |
168 | savesegment(gs, gs); |
169 | } else { | |
65ea5b03 | 170 | sp = (unsigned long) (®s->sp); |
9d975ebd PE |
171 | savesegment(ss, ss); |
172 | savesegment(gs, gs); | |
173 | } | |
1da177e4 LT |
174 | |
175 | printk("\n"); | |
60812a4a LT |
176 | printk("Pid: %d, comm: %s %s (%s %.*s)\n", |
177 | task_pid_nr(current), current->comm, | |
9d975ebd PE |
178 | print_tainted(), init_utsname()->release, |
179 | (int)strcspn(init_utsname()->version, " "), | |
180 | init_utsname()->version); | |
181 | ||
182 | printk("EIP: %04x:[<%08lx>] EFLAGS: %08lx CPU: %d\n", | |
92bc2056 | 183 | (u16)regs->cs, regs->ip, regs->flags, |
9d975ebd | 184 | smp_processor_id()); |
65ea5b03 | 185 | print_symbol("EIP is at %s\n", regs->ip); |
1da177e4 | 186 | |
1da177e4 | 187 | printk("EAX: %08lx EBX: %08lx ECX: %08lx EDX: %08lx\n", |
65ea5b03 | 188 | regs->ax, regs->bx, regs->cx, regs->dx); |
9d975ebd | 189 | printk("ESI: %08lx EDI: %08lx EBP: %08lx ESP: %08lx\n", |
65ea5b03 | 190 | regs->si, regs->di, regs->bp, sp); |
9d975ebd | 191 | printk(" DS: %04x ES: %04x FS: %04x GS: %04x SS: %04x\n", |
92bc2056 | 192 | (u16)regs->ds, (u16)regs->es, (u16)regs->fs, gs, ss); |
9d975ebd PE |
193 | |
194 | if (!all) | |
195 | return; | |
1da177e4 | 196 | |
4bb0d3ec ZA |
197 | cr0 = read_cr0(); |
198 | cr2 = read_cr2(); | |
199 | cr3 = read_cr3(); | |
ff6e8c0d | 200 | cr4 = read_cr4_safe(); |
9d975ebd PE |
201 | printk("CR0: %08lx CR2: %08lx CR3: %08lx CR4: %08lx\n", |
202 | cr0, cr2, cr3, cr4); | |
bb1995d5 AS |
203 | |
204 | get_debugreg(d0, 0); | |
205 | get_debugreg(d1, 1); | |
206 | get_debugreg(d2, 2); | |
207 | get_debugreg(d3, 3); | |
208 | printk("DR0: %08lx DR1: %08lx DR2: %08lx DR3: %08lx\n", | |
209 | d0, d1, d2, d3); | |
9d975ebd | 210 | |
bb1995d5 AS |
211 | get_debugreg(d6, 6); |
212 | get_debugreg(d7, 7); | |
9d975ebd PE |
213 | printk("DR6: %08lx DR7: %08lx\n", |
214 | d6, d7); | |
215 | } | |
bb1995d5 | 216 | |
9d975ebd PE |
217 | void show_regs(struct pt_regs *regs) |
218 | { | |
219 | __show_registers(regs, 1); | |
5bc27dc2 | 220 | show_trace(NULL, regs, ®s->sp, regs->bp); |
1da177e4 LT |
221 | } |
222 | ||
223 | /* | |
65ea5b03 PA |
224 | * This gets run with %bx containing the |
225 | * function to call, and %dx containing | |
1da177e4 LT |
226 | * the "args". |
227 | */ | |
228 | extern void kernel_thread_helper(void); | |
1da177e4 LT |
229 | |
230 | /* | |
231 | * Create a kernel thread | |
232 | */ | |
233 | int kernel_thread(int (*fn)(void *), void * arg, unsigned long flags) | |
234 | { | |
235 | struct pt_regs regs; | |
236 | ||
237 | memset(®s, 0, sizeof(regs)); | |
238 | ||
65ea5b03 PA |
239 | regs.bx = (unsigned long) fn; |
240 | regs.dx = (unsigned long) arg; | |
1da177e4 | 241 | |
65ea5b03 PA |
242 | regs.ds = __USER_DS; |
243 | regs.es = __USER_DS; | |
244 | regs.fs = __KERNEL_PERCPU; | |
245 | regs.orig_ax = -1; | |
246 | regs.ip = (unsigned long) kernel_thread_helper; | |
247 | regs.cs = __KERNEL_CS | get_kernel_rpl(); | |
248 | regs.flags = X86_EFLAGS_IF | X86_EFLAGS_SF | X86_EFLAGS_PF | 0x2; | |
1da177e4 LT |
249 | |
250 | /* Ok, create the new process.. */ | |
8cf2c519 | 251 | return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, ®s, 0, NULL, NULL); |
1da177e4 | 252 | } |
129f6946 | 253 | EXPORT_SYMBOL(kernel_thread); |
1da177e4 LT |
254 | |
255 | /* | |
256 | * Free current thread data structures etc.. | |
257 | */ | |
258 | void exit_thread(void) | |
259 | { | |
1da177e4 | 260 | /* The process may have allocated an io port bitmap... nuke it. */ |
b3cf2576 SE |
261 | if (unlikely(test_thread_flag(TIF_IO_BITMAP))) { |
262 | struct task_struct *tsk = current; | |
263 | struct thread_struct *t = &tsk->thread; | |
1da177e4 LT |
264 | int cpu = get_cpu(); |
265 | struct tss_struct *tss = &per_cpu(init_tss, cpu); | |
266 | ||
267 | kfree(t->io_bitmap_ptr); | |
268 | t->io_bitmap_ptr = NULL; | |
b3cf2576 | 269 | clear_thread_flag(TIF_IO_BITMAP); |
1da177e4 LT |
270 | /* |
271 | * Careful, clear this in the TSS too: | |
272 | */ | |
273 | memset(tss->io_bitmap, 0xff, tss->io_bitmap_max); | |
274 | t->io_bitmap_max = 0; | |
275 | tss->io_bitmap_owner = NULL; | |
276 | tss->io_bitmap_max = 0; | |
a75c54f9 | 277 | tss->x86_tss.io_bitmap_base = INVALID_IO_BITMAP_OFFSET; |
1da177e4 LT |
278 | put_cpu(); |
279 | } | |
280 | } | |
281 | ||
282 | void flush_thread(void) | |
283 | { | |
284 | struct task_struct *tsk = current; | |
285 | ||
0f534093 RM |
286 | tsk->thread.debugreg0 = 0; |
287 | tsk->thread.debugreg1 = 0; | |
288 | tsk->thread.debugreg2 = 0; | |
289 | tsk->thread.debugreg3 = 0; | |
290 | tsk->thread.debugreg6 = 0; | |
291 | tsk->thread.debugreg7 = 0; | |
1da177e4 | 292 | memset(tsk->thread.tls_array, 0, sizeof(tsk->thread.tls_array)); |
b3cf2576 | 293 | clear_tsk_thread_flag(tsk, TIF_DEBUG); |
1da177e4 LT |
294 | /* |
295 | * Forget coprocessor state.. | |
296 | */ | |
75118a82 | 297 | tsk->fpu_counter = 0; |
1da177e4 LT |
298 | clear_fpu(tsk); |
299 | clear_used_math(); | |
300 | } | |
301 | ||
302 | void release_thread(struct task_struct *dead_task) | |
303 | { | |
2684927c | 304 | BUG_ON(dead_task->mm); |
1da177e4 LT |
305 | release_vm86_irqs(dead_task); |
306 | } | |
307 | ||
308 | /* | |
309 | * This gets called before we allocate a new thread and copy | |
310 | * the current task into it. | |
311 | */ | |
312 | void prepare_to_copy(struct task_struct *tsk) | |
313 | { | |
314 | unlazy_fpu(tsk); | |
315 | } | |
316 | ||
65ea5b03 | 317 | int copy_thread(int nr, unsigned long clone_flags, unsigned long sp, |
1da177e4 LT |
318 | unsigned long unused, |
319 | struct task_struct * p, struct pt_regs * regs) | |
320 | { | |
321 | struct pt_regs * childregs; | |
322 | struct task_struct *tsk; | |
323 | int err; | |
324 | ||
07b047fc | 325 | childregs = task_pt_regs(p); |
f48d9663 | 326 | *childregs = *regs; |
65ea5b03 PA |
327 | childregs->ax = 0; |
328 | childregs->sp = sp; | |
f48d9663 | 329 | |
faca6227 PA |
330 | p->thread.sp = (unsigned long) childregs; |
331 | p->thread.sp0 = (unsigned long) (childregs+1); | |
1da177e4 | 332 | |
faca6227 | 333 | p->thread.ip = (unsigned long) ret_from_fork; |
1da177e4 | 334 | |
6612538c | 335 | savesegment(gs, p->thread.gs); |
1da177e4 LT |
336 | |
337 | tsk = current; | |
b3cf2576 | 338 | if (unlikely(test_tsk_thread_flag(tsk, TIF_IO_BITMAP))) { |
52978be6 AD |
339 | p->thread.io_bitmap_ptr = kmemdup(tsk->thread.io_bitmap_ptr, |
340 | IO_BITMAP_BYTES, GFP_KERNEL); | |
1da177e4 LT |
341 | if (!p->thread.io_bitmap_ptr) { |
342 | p->thread.io_bitmap_max = 0; | |
343 | return -ENOMEM; | |
344 | } | |
b3cf2576 | 345 | set_tsk_thread_flag(p, TIF_IO_BITMAP); |
1da177e4 LT |
346 | } |
347 | ||
efd1ca52 RM |
348 | err = 0; |
349 | ||
1da177e4 LT |
350 | /* |
351 | * Set a new TLS for the child thread? | |
352 | */ | |
efd1ca52 RM |
353 | if (clone_flags & CLONE_SETTLS) |
354 | err = do_set_thread_area(p, -1, | |
65ea5b03 | 355 | (struct user_desc __user *)childregs->si, 0); |
1da177e4 | 356 | |
1da177e4 LT |
357 | if (err && p->thread.io_bitmap_ptr) { |
358 | kfree(p->thread.io_bitmap_ptr); | |
359 | p->thread.io_bitmap_max = 0; | |
360 | } | |
361 | return err; | |
362 | } | |
363 | ||
513ad84b IM |
364 | void |
365 | start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp) | |
366 | { | |
367 | __asm__("movl %0, %%gs" :: "r"(0)); | |
368 | regs->fs = 0; | |
369 | set_fs(USER_DS); | |
370 | regs->ds = __USER_DS; | |
371 | regs->es = __USER_DS; | |
372 | regs->ss = __USER_DS; | |
373 | regs->cs = __USER_CS; | |
374 | regs->ip = new_ip; | |
375 | regs->sp = new_sp; | |
aa283f49 SS |
376 | /* |
377 | * Free the old FP and other extended state | |
378 | */ | |
379 | free_thread_xstate(current); | |
513ad84b IM |
380 | } |
381 | EXPORT_SYMBOL_GPL(start_thread); | |
382 | ||
bdb4f156 | 383 | static void hard_disable_TSC(void) |
cf99abac AA |
384 | { |
385 | write_cr4(read_cr4() | X86_CR4_TSD); | |
386 | } | |
529e25f6 | 387 | |
cf99abac AA |
388 | void disable_TSC(void) |
389 | { | |
390 | preempt_disable(); | |
391 | if (!test_and_set_thread_flag(TIF_NOTSC)) | |
392 | /* | |
393 | * Must flip the CPU state synchronously with | |
394 | * TIF_NOTSC in the current running context. | |
395 | */ | |
396 | hard_disable_TSC(); | |
397 | preempt_enable(); | |
398 | } | |
529e25f6 | 399 | |
bdb4f156 | 400 | static void hard_enable_TSC(void) |
cf99abac AA |
401 | { |
402 | write_cr4(read_cr4() & ~X86_CR4_TSD); | |
403 | } | |
529e25f6 | 404 | |
a4928cff | 405 | static void enable_TSC(void) |
529e25f6 EB |
406 | { |
407 | preempt_disable(); | |
408 | if (test_and_clear_thread_flag(TIF_NOTSC)) | |
409 | /* | |
410 | * Must flip the CPU state synchronously with | |
411 | * TIF_NOTSC in the current running context. | |
412 | */ | |
413 | hard_enable_TSC(); | |
414 | preempt_enable(); | |
415 | } | |
416 | ||
417 | int get_tsc_mode(unsigned long adr) | |
418 | { | |
419 | unsigned int val; | |
420 | ||
421 | if (test_thread_flag(TIF_NOTSC)) | |
422 | val = PR_TSC_SIGSEGV; | |
423 | else | |
424 | val = PR_TSC_ENABLE; | |
425 | ||
426 | return put_user(val, (unsigned int __user *)adr); | |
427 | } | |
428 | ||
429 | int set_tsc_mode(unsigned int val) | |
430 | { | |
431 | if (val == PR_TSC_SIGSEGV) | |
432 | disable_TSC(); | |
433 | else if (val == PR_TSC_ENABLE) | |
434 | enable_TSC(); | |
435 | else | |
436 | return -EINVAL; | |
437 | ||
438 | return 0; | |
439 | } | |
cf99abac AA |
440 | |
441 | static noinline void | |
442 | __switch_to_xtra(struct task_struct *prev_p, struct task_struct *next_p, | |
443 | struct tss_struct *tss) | |
1da177e4 | 444 | { |
7e991604 | 445 | struct thread_struct *prev, *next; |
eee3af4a | 446 | unsigned long debugctl; |
b3cf2576 | 447 | |
7e991604 | 448 | prev = &prev_p->thread; |
b3cf2576 SE |
449 | next = &next_p->thread; |
450 | ||
eee3af4a MM |
451 | debugctl = prev->debugctlmsr; |
452 | if (next->ds_area_msr != prev->ds_area_msr) { | |
453 | /* we clear debugctl to make sure DS | |
454 | * is not in use when we change it */ | |
455 | debugctl = 0; | |
5b0e5084 | 456 | update_debugctlmsr(0); |
eee3af4a MM |
457 | wrmsr(MSR_IA32_DS_AREA, next->ds_area_msr, 0); |
458 | } | |
459 | ||
460 | if (next->debugctlmsr != debugctl) | |
5b0e5084 | 461 | update_debugctlmsr(next->debugctlmsr); |
7e991604 | 462 | |
b3cf2576 | 463 | if (test_tsk_thread_flag(next_p, TIF_DEBUG)) { |
0f534093 RM |
464 | set_debugreg(next->debugreg0, 0); |
465 | set_debugreg(next->debugreg1, 1); | |
466 | set_debugreg(next->debugreg2, 2); | |
467 | set_debugreg(next->debugreg3, 3); | |
b3cf2576 | 468 | /* no 4 and 5 */ |
0f534093 RM |
469 | set_debugreg(next->debugreg6, 6); |
470 | set_debugreg(next->debugreg7, 7); | |
b3cf2576 SE |
471 | } |
472 | ||
cf99abac AA |
473 | if (test_tsk_thread_flag(prev_p, TIF_NOTSC) ^ |
474 | test_tsk_thread_flag(next_p, TIF_NOTSC)) { | |
475 | /* prev and next are different */ | |
476 | if (test_tsk_thread_flag(next_p, TIF_NOTSC)) | |
477 | hard_disable_TSC(); | |
478 | else | |
479 | hard_enable_TSC(); | |
480 | } | |
cf99abac | 481 | |
b4ef95de | 482 | #ifdef X86_BTS |
eee3af4a MM |
483 | if (test_tsk_thread_flag(prev_p, TIF_BTS_TRACE_TS)) |
484 | ptrace_bts_take_timestamp(prev_p, BTS_TASK_DEPARTS); | |
485 | ||
486 | if (test_tsk_thread_flag(next_p, TIF_BTS_TRACE_TS)) | |
487 | ptrace_bts_take_timestamp(next_p, BTS_TASK_ARRIVES); | |
b4ef95de | 488 | #endif |
eee3af4a MM |
489 | |
490 | ||
b3cf2576 | 491 | if (!test_tsk_thread_flag(next_p, TIF_IO_BITMAP)) { |
1da177e4 LT |
492 | /* |
493 | * Disable the bitmap via an invalid offset. We still cache | |
494 | * the previous bitmap owner and the IO bitmap contents: | |
495 | */ | |
a75c54f9 | 496 | tss->x86_tss.io_bitmap_base = INVALID_IO_BITMAP_OFFSET; |
1da177e4 LT |
497 | return; |
498 | } | |
b3cf2576 | 499 | |
1da177e4 LT |
500 | if (likely(next == tss->io_bitmap_owner)) { |
501 | /* | |
502 | * Previous owner of the bitmap (hence the bitmap content) | |
503 | * matches the next task, we dont have to do anything but | |
504 | * to set a valid offset in the TSS: | |
505 | */ | |
a75c54f9 | 506 | tss->x86_tss.io_bitmap_base = IO_BITMAP_OFFSET; |
1da177e4 LT |
507 | return; |
508 | } | |
509 | /* | |
510 | * Lazy TSS's I/O bitmap copy. We set an invalid offset here | |
511 | * and we let the task to get a GPF in case an I/O instruction | |
512 | * is performed. The handler of the GPF will verify that the | |
513 | * faulting task has a valid I/O bitmap and, it true, does the | |
514 | * real copy and restart the instruction. This will save us | |
515 | * redundant copies when the currently switched task does not | |
516 | * perform any I/O during its timeslice. | |
517 | */ | |
a75c54f9 | 518 | tss->x86_tss.io_bitmap_base = INVALID_IO_BITMAP_OFFSET_LAZY; |
1da177e4 | 519 | } |
1da177e4 LT |
520 | |
521 | /* | |
522 | * switch_to(x,yn) should switch tasks from x to y. | |
523 | * | |
524 | * We fsave/fwait so that an exception goes off at the right time | |
525 | * (as a call from the fsave or fwait in effect) rather than to | |
526 | * the wrong process. Lazy FP saving no longer makes any sense | |
527 | * with modern CPU's, and this simplifies a lot of things (SMP | |
528 | * and UP become the same). | |
529 | * | |
530 | * NOTE! We used to use the x86 hardware context switching. The | |
531 | * reason for not using it any more becomes apparent when you | |
532 | * try to recover gracefully from saved state that is no longer | |
533 | * valid (stale segment register values in particular). With the | |
534 | * hardware task-switch, there is no way to fix up bad state in | |
535 | * a reasonable manner. | |
536 | * | |
537 | * The fact that Intel documents the hardware task-switching to | |
538 | * be slow is a fairly red herring - this code is not noticeably | |
539 | * faster. However, there _is_ some room for improvement here, | |
540 | * so the performance issues may eventually be a valid point. | |
541 | * More important, however, is the fact that this allows us much | |
542 | * more flexibility. | |
543 | * | |
65ea5b03 | 544 | * The return value (in %ax) will be the "prev" task after |
1da177e4 LT |
545 | * the task-switch, and shows up in ret_from_fork in entry.S, |
546 | * for example. | |
547 | */ | |
75604d7f | 548 | struct task_struct * __switch_to(struct task_struct *prev_p, struct task_struct *next_p) |
1da177e4 LT |
549 | { |
550 | struct thread_struct *prev = &prev_p->thread, | |
551 | *next = &next_p->thread; | |
552 | int cpu = smp_processor_id(); | |
553 | struct tss_struct *tss = &per_cpu(init_tss, cpu); | |
554 | ||
555 | /* never put a printk in __switch_to... printk() calls wake_up*() indirectly */ | |
556 | ||
557 | __unlazy_fpu(prev_p); | |
558 | ||
acc20761 CE |
559 | |
560 | /* we're going to use this soon, after a few expensive things */ | |
561 | if (next_p->fpu_counter > 5) | |
61c4628b | 562 | prefetch(next->xstate); |
acc20761 | 563 | |
1da177e4 | 564 | /* |
e7a2ff59 | 565 | * Reload esp0. |
1da177e4 | 566 | */ |
faca6227 | 567 | load_sp0(tss, next); |
1da177e4 LT |
568 | |
569 | /* | |
464d1a78 | 570 | * Save away %gs. No need to save %fs, as it was saved on the |
f95d47ca JF |
571 | * stack on entry. No need to save %es and %ds, as those are |
572 | * always kernel segments while inside the kernel. Doing this | |
573 | * before setting the new TLS descriptors avoids the situation | |
574 | * where we temporarily have non-reloadable segments in %fs | |
575 | * and %gs. This could be an issue if the NMI handler ever | |
576 | * used %fs or %gs (it does not today), or if the kernel is | |
577 | * running inside of a hypervisor layer. | |
1da177e4 | 578 | */ |
464d1a78 | 579 | savesegment(gs, prev->gs); |
1da177e4 LT |
580 | |
581 | /* | |
e7a2ff59 | 582 | * Load the per-thread Thread-Local Storage descriptor. |
1da177e4 | 583 | */ |
e7a2ff59 | 584 | load_TLS(next, cpu); |
1da177e4 | 585 | |
8b151144 ZA |
586 | /* |
587 | * Restore IOPL if needed. In normal use, the flags restore | |
588 | * in the switch assembly will handle this. But if the kernel | |
589 | * is running virtualized at a non-zero CPL, the popf will | |
590 | * not restore flags, so it must be done in a separate step. | |
591 | */ | |
592 | if (get_kernel_rpl() && unlikely(prev->iopl != next->iopl)) | |
593 | set_iopl_mask(next->iopl); | |
594 | ||
1da177e4 | 595 | /* |
b3cf2576 | 596 | * Now maybe handle debug registers and/or IO bitmaps |
1da177e4 | 597 | */ |
cf99abac AA |
598 | if (unlikely(task_thread_info(prev_p)->flags & _TIF_WORK_CTXSW_PREV || |
599 | task_thread_info(next_p)->flags & _TIF_WORK_CTXSW_NEXT)) | |
600 | __switch_to_xtra(prev_p, next_p, tss); | |
ffaa8bd6 | 601 | |
9226d125 ZA |
602 | /* |
603 | * Leave lazy mode, flushing any hypercalls made here. | |
604 | * This must be done before restoring TLS segments so | |
605 | * the GDT and LDT are properly updated, and must be | |
606 | * done before math_state_restore, so the TS bit is up | |
607 | * to date. | |
608 | */ | |
609 | arch_leave_lazy_cpu_mode(); | |
610 | ||
acc20761 CE |
611 | /* If the task has used fpu the last 5 timeslices, just do a full |
612 | * restore of the math state immediately to avoid the trap; the | |
613 | * chances of needing FPU soon are obviously high now | |
870568b3 SS |
614 | * |
615 | * tsk_used_math() checks prevent calling math_state_restore(), | |
616 | * which can sleep in the case of !tsk_used_math() | |
acc20761 | 617 | */ |
870568b3 | 618 | if (tsk_used_math(next_p) && next_p->fpu_counter > 5) |
acc20761 CE |
619 | math_state_restore(); |
620 | ||
9226d125 ZA |
621 | /* |
622 | * Restore %gs if needed (which is common) | |
623 | */ | |
624 | if (prev->gs | next->gs) | |
625 | loadsegment(gs, next->gs); | |
626 | ||
7c3576d2 | 627 | x86_write_percpu(current_task, next_p); |
9226d125 | 628 | |
1da177e4 LT |
629 | return prev_p; |
630 | } | |
631 | ||
632 | asmlinkage int sys_fork(struct pt_regs regs) | |
633 | { | |
65ea5b03 | 634 | return do_fork(SIGCHLD, regs.sp, ®s, 0, NULL, NULL); |
1da177e4 LT |
635 | } |
636 | ||
637 | asmlinkage int sys_clone(struct pt_regs regs) | |
638 | { | |
639 | unsigned long clone_flags; | |
640 | unsigned long newsp; | |
641 | int __user *parent_tidptr, *child_tidptr; | |
642 | ||
65ea5b03 PA |
643 | clone_flags = regs.bx; |
644 | newsp = regs.cx; | |
645 | parent_tidptr = (int __user *)regs.dx; | |
646 | child_tidptr = (int __user *)regs.di; | |
1da177e4 | 647 | if (!newsp) |
65ea5b03 | 648 | newsp = regs.sp; |
1da177e4 LT |
649 | return do_fork(clone_flags, newsp, ®s, 0, parent_tidptr, child_tidptr); |
650 | } | |
651 | ||
652 | /* | |
653 | * This is trivial, and on the face of it looks like it | |
654 | * could equally well be done in user mode. | |
655 | * | |
656 | * Not so, for quite unobvious reasons - register pressure. | |
657 | * In user mode vfork() cannot have a stack frame, and if | |
658 | * done by calling the "clone()" system call directly, you | |
659 | * do not have enough call-clobbered registers to hold all | |
660 | * the information you need. | |
661 | */ | |
662 | asmlinkage int sys_vfork(struct pt_regs regs) | |
663 | { | |
65ea5b03 | 664 | return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, regs.sp, ®s, 0, NULL, NULL); |
1da177e4 LT |
665 | } |
666 | ||
667 | /* | |
668 | * sys_execve() executes a new program. | |
669 | */ | |
670 | asmlinkage int sys_execve(struct pt_regs regs) | |
671 | { | |
672 | int error; | |
673 | char * filename; | |
674 | ||
65ea5b03 | 675 | filename = getname((char __user *) regs.bx); |
1da177e4 LT |
676 | error = PTR_ERR(filename); |
677 | if (IS_ERR(filename)) | |
678 | goto out; | |
679 | error = do_execve(filename, | |
65ea5b03 PA |
680 | (char __user * __user *) regs.cx, |
681 | (char __user * __user *) regs.dx, | |
1da177e4 LT |
682 | ®s); |
683 | if (error == 0) { | |
1da177e4 LT |
684 | /* Make sure we don't return using sysenter.. */ |
685 | set_thread_flag(TIF_IRET); | |
686 | } | |
687 | putname(filename); | |
688 | out: | |
689 | return error; | |
690 | } | |
691 | ||
692 | #define top_esp (THREAD_SIZE - sizeof(unsigned long)) | |
693 | #define top_ebp (THREAD_SIZE - 2*sizeof(unsigned long)) | |
694 | ||
695 | unsigned long get_wchan(struct task_struct *p) | |
696 | { | |
65ea5b03 | 697 | unsigned long bp, sp, ip; |
1da177e4 LT |
698 | unsigned long stack_page; |
699 | int count = 0; | |
700 | if (!p || p == current || p->state == TASK_RUNNING) | |
701 | return 0; | |
65e0fdff | 702 | stack_page = (unsigned long)task_stack_page(p); |
faca6227 | 703 | sp = p->thread.sp; |
65ea5b03 | 704 | if (!stack_page || sp < stack_page || sp > top_esp+stack_page) |
1da177e4 | 705 | return 0; |
65ea5b03 PA |
706 | /* include/asm-i386/system.h:switch_to() pushes bp last. */ |
707 | bp = *(unsigned long *) sp; | |
1da177e4 | 708 | do { |
65ea5b03 | 709 | if (bp < stack_page || bp > top_ebp+stack_page) |
1da177e4 | 710 | return 0; |
65ea5b03 PA |
711 | ip = *(unsigned long *) (bp+4); |
712 | if (!in_sched_functions(ip)) | |
713 | return ip; | |
714 | bp = *(unsigned long *) bp; | |
1da177e4 LT |
715 | } while (count++ < 16); |
716 | return 0; | |
717 | } | |
718 | ||
1da177e4 LT |
719 | unsigned long arch_align_stack(unsigned long sp) |
720 | { | |
c16b63e0 | 721 | if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space) |
1da177e4 LT |
722 | sp -= get_random_int() % 8192; |
723 | return sp & ~0xf; | |
724 | } | |
c1d171a0 JK |
725 | |
726 | unsigned long arch_randomize_brk(struct mm_struct *mm) | |
727 | { | |
728 | unsigned long range_end = mm->brk + 0x02000000; | |
729 | return randomize_range(mm->brk, range_end, 0) ? : mm->brk; | |
730 | } |