]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - arch/x86/kernel/process_32.c
generic, x86: add prctl commands PR_GET_TSC and PR_SET_TSC
[mirror_ubuntu-artful-kernel.git] / arch / x86 / kernel / process_32.c
CommitLineData
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>
1da177e4
LT
39
40#include <asm/uaccess.h>
41#include <asm/pgtable.h>
42#include <asm/system.h>
43#include <asm/io.h>
44#include <asm/ldt.h>
45#include <asm/processor.h>
46#include <asm/i387.h>
1da177e4
LT
47#include <asm/desc.h>
48#ifdef CONFIG_MATH_EMULATION
49#include <asm/math_emu.h>
50#endif
51
1da177e4
LT
52#include <linux/err.h>
53
f3705136
ZM
54#include <asm/tlbflush.h>
55#include <asm/cpu.h>
718fc13b 56#include <asm/kdebug.h>
f3705136 57
1da177e4
LT
58asmlinkage void ret_from_fork(void) __asm__("ret_from_fork");
59
60static int hlt_counter;
61
62unsigned long boot_option_idle_override = 0;
63EXPORT_SYMBOL(boot_option_idle_override);
64
7c3576d2
JF
65DEFINE_PER_CPU(struct task_struct *, current_task) = &init_task;
66EXPORT_PER_CPU_SYMBOL(current_task);
67
68DEFINE_PER_CPU(int, cpu_number);
69EXPORT_PER_CPU_SYMBOL(cpu_number);
70
1da177e4
LT
71/*
72 * Return saved PC of a blocked thread.
73 */
74unsigned long thread_saved_pc(struct task_struct *tsk)
75{
faca6227 76 return ((unsigned long *)tsk->thread.sp)[3];
1da177e4
LT
77}
78
79/*
80 * Powermanagement idle function, if any..
81 */
82void (*pm_idle)(void);
129f6946 83EXPORT_SYMBOL(pm_idle);
1da177e4
LT
84
85void disable_hlt(void)
86{
87 hlt_counter++;
88}
89
90EXPORT_SYMBOL(disable_hlt);
91
92void enable_hlt(void)
93{
94 hlt_counter--;
95}
96
97EXPORT_SYMBOL(enable_hlt);
98
99/*
100 * We use this if we don't have any better
101 * idle routine..
102 */
103void default_idle(void)
104{
105 if (!hlt_counter && boot_cpu_data.hlt_works_ok) {
495ab9c0 106 current_thread_info()->status &= ~TS_POLLING;
0888f06a
IM
107 /*
108 * TS_POLLING-cleared state must be visible before we
109 * test NEED_RESCHED:
110 */
111 smp_mb();
112
72690a21 113 local_irq_disable();
5ee613b6 114 if (!need_resched()) {
72690a21 115 safe_halt(); /* enables interrupts racelessly */
5ee613b6 116 local_irq_disable();
5ee613b6
IM
117 }
118 local_irq_enable();
495ab9c0 119 current_thread_info()->status |= TS_POLLING;
1da177e4 120 } else {
3b22ec7b 121 local_irq_enable();
72690a21
AK
122 /* loop is done by the caller */
123 cpu_relax();
1da177e4
LT
124 }
125}
129f6946
AD
126#ifdef CONFIG_APM_MODULE
127EXPORT_SYMBOL(default_idle);
128#endif
1da177e4
LT
129
130/*
131 * On SMP it's slightly faster (but much more power-consuming!)
132 * to poll the ->work.need_resched flag instead of waiting for the
133 * cross-CPU IPI to arrive. Use this option with caution.
134 */
6612538c 135static void poll_idle(void)
1da177e4 136{
3b22ec7b 137 local_irq_enable();
72690a21 138 cpu_relax();
1da177e4
LT
139}
140
f3705136
ZM
141#ifdef CONFIG_HOTPLUG_CPU
142#include <asm/nmi.h>
143/* We don't actually take CPU down, just spin without interrupts. */
144static inline void play_dead(void)
145{
e1367daf
LS
146 /* This must be done before dead CPU ack */
147 cpu_exit_clear();
148 wbinvd();
149 mb();
f3705136
ZM
150 /* Ack it */
151 __get_cpu_var(cpu_state) = CPU_DEAD;
152
e1367daf
LS
153 /*
154 * With physical CPU hotplug, we should halt the cpu
155 */
f3705136 156 local_irq_disable();
e1367daf 157 while (1)
f2ab4461 158 halt();
f3705136
ZM
159}
160#else
161static inline void play_dead(void)
162{
163 BUG();
164}
165#endif /* CONFIG_HOTPLUG_CPU */
166
1da177e4
LT
167/*
168 * The idle thread. There's no useful work to be
169 * done, so just try to conserve power and have a
170 * low exit latency (ie sit in a loop waiting for
171 * somebody to say that they'd like to reschedule)
172 */
f3705136 173void cpu_idle(void)
1da177e4 174{
5bfb5d69 175 int cpu = smp_processor_id();
f3705136 176
495ab9c0 177 current_thread_info()->status |= TS_POLLING;
64c7c8f8 178
1da177e4
LT
179 /* endless idle loop with no priority at all */
180 while (1) {
74167347 181 tick_nohz_stop_sched_tick();
1da177e4
LT
182 while (!need_resched()) {
183 void (*idle)(void);
184
f1d1a842 185 check_pgt_cache();
1da177e4
LT
186 rmb();
187 idle = pm_idle;
188
0723a69a
BL
189 if (rcu_pending(cpu))
190 rcu_check_callbacks(cpu, 0);
191
1da177e4
LT
192 if (!idle)
193 idle = default_idle;
194
f3705136
ZM
195 if (cpu_is_offline(cpu))
196 play_dead();
197
1da177e4
LT
198 __get_cpu_var(irq_stat).idle_timestamp = jiffies;
199 idle();
200 }
74167347 201 tick_nohz_restart_sched_tick();
5bfb5d69 202 preempt_enable_no_resched();
1da177e4 203 schedule();
5bfb5d69 204 preempt_disable();
1da177e4
LT
205 }
206}
207
40d6a146
SR
208static void do_nothing(void *unused)
209{
210}
211
783e391b
VP
212/*
213 * cpu_idle_wait - Used to ensure that all the CPUs discard old value of
214 * pm_idle and update to new pm_idle value. Required while changing pm_idle
215 * handler on SMP systems.
216 *
217 * Caller must have changed pm_idle to the new value before the call. Old
218 * pm_idle value will not be used by any CPU after the return of this function.
219 */
1da177e4
LT
220void cpu_idle_wait(void)
221{
783e391b
VP
222 smp_mb();
223 /* kick all the CPUs so that they exit out of pm_idle */
224 smp_call_function(do_nothing, NULL, 0, 1);
1da177e4
LT
225}
226EXPORT_SYMBOL_GPL(cpu_idle_wait);
227
228/*
229 * This uses new MONITOR/MWAIT instructions on P4 processors with PNI,
230 * which can obviate IPI to trigger checking of need_resched.
231 * We execute MONITOR against need_resched and enter optimized wait state
232 * through MWAIT. Whenever someone changes need_resched, we would be woken
233 * up from MWAIT (without an IPI).
991528d7
VP
234 *
235 * New with Core Duo processors, MWAIT can take some hints based on CPU
236 * capability.
1da177e4 237 */
65ea5b03 238void mwait_idle_with_hints(unsigned long ax, unsigned long cx)
1da177e4 239{
991528d7 240 if (!need_resched()) {
64c7c8f8
NP
241 __monitor((void *)&current_thread_info()->flags, 0, 0);
242 smp_mb();
991528d7 243 if (!need_resched())
3b22ec7b
GOC
244 __sti_mwait(ax, cx);
245 else
246 local_irq_enable();
247 } else
248 local_irq_enable();
1da177e4
LT
249}
250
991528d7
VP
251/* Default MONITOR/MWAIT with no hints, used for default C1 state */
252static void mwait_idle(void)
253{
254 local_irq_enable();
72690a21 255 mwait_idle_with_hints(0, 0);
991528d7
VP
256}
257
4c02ad1e 258static int __cpuinit mwait_usable(const struct cpuinfo_x86 *c)
0c07ee38
AK
259{
260 if (force_mwait)
261 return 1;
262 /* Any C1 states supported? */
263 return c->cpuid_level >= 5 && ((cpuid_edx(5) >> 4) & 0xf) > 0;
264}
265
3446fa05 266void __cpuinit select_idle_routine(const struct cpuinfo_x86 *c)
1da177e4 267{
27415a4f
HS
268 static int selected;
269
270 if (selected)
271 return;
272#ifdef CONFIG_X86_SMP
273 if (pm_idle == poll_idle && smp_num_siblings > 1) {
274 printk(KERN_WARNING "WARNING: polling idle and HT enabled,"
275 " performance may degrade.\n");
276 }
277#endif
0c07ee38 278 if (cpu_has(c, X86_FEATURE_MWAIT) && mwait_usable(c)) {
1da177e4
LT
279 /*
280 * Skip, if setup has overridden idle.
281 * One CPU supports mwait => All CPUs supports mwait
282 */
283 if (!pm_idle) {
27415a4f 284 printk(KERN_INFO "using mwait in idle threads.\n");
1da177e4
LT
285 pm_idle = mwait_idle;
286 }
287 }
27415a4f 288 selected = 1;
1da177e4
LT
289}
290
f039b754 291static int __init idle_setup(char *str)
1da177e4 292{
f039b754 293 if (!strcmp(str, "poll")) {
1da177e4
LT
294 printk("using polling idle threads.\n");
295 pm_idle = poll_idle;
f039b754
AK
296 } else if (!strcmp(str, "mwait"))
297 force_mwait = 1;
298 else
299 return -1;
1da177e4
LT
300
301 boot_option_idle_override = 1;
f039b754 302 return 0;
1da177e4 303}
f039b754 304early_param("idle", idle_setup);
1da177e4 305
9d975ebd 306void __show_registers(struct pt_regs *regs, int all)
1da177e4
LT
307{
308 unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L;
bb1995d5 309 unsigned long d0, d1, d2, d3, d6, d7;
65ea5b03 310 unsigned long sp;
9d975ebd
PE
311 unsigned short ss, gs;
312
313 if (user_mode_vm(regs)) {
65ea5b03
PA
314 sp = regs->sp;
315 ss = regs->ss & 0xffff;
9d975ebd
PE
316 savesegment(gs, gs);
317 } else {
65ea5b03 318 sp = (unsigned long) (&regs->sp);
9d975ebd
PE
319 savesegment(ss, ss);
320 savesegment(gs, gs);
321 }
1da177e4
LT
322
323 printk("\n");
60812a4a
LT
324 printk("Pid: %d, comm: %s %s (%s %.*s)\n",
325 task_pid_nr(current), current->comm,
9d975ebd
PE
326 print_tainted(), init_utsname()->release,
327 (int)strcspn(init_utsname()->version, " "),
328 init_utsname()->version);
329
330 printk("EIP: %04x:[<%08lx>] EFLAGS: %08lx CPU: %d\n",
92bc2056 331 (u16)regs->cs, regs->ip, regs->flags,
9d975ebd 332 smp_processor_id());
65ea5b03 333 print_symbol("EIP is at %s\n", regs->ip);
1da177e4 334
1da177e4 335 printk("EAX: %08lx EBX: %08lx ECX: %08lx EDX: %08lx\n",
65ea5b03 336 regs->ax, regs->bx, regs->cx, regs->dx);
9d975ebd 337 printk("ESI: %08lx EDI: %08lx EBP: %08lx ESP: %08lx\n",
65ea5b03 338 regs->si, regs->di, regs->bp, sp);
9d975ebd 339 printk(" DS: %04x ES: %04x FS: %04x GS: %04x SS: %04x\n",
92bc2056 340 (u16)regs->ds, (u16)regs->es, (u16)regs->fs, gs, ss);
9d975ebd
PE
341
342 if (!all)
343 return;
1da177e4 344
4bb0d3ec
ZA
345 cr0 = read_cr0();
346 cr2 = read_cr2();
347 cr3 = read_cr3();
ff6e8c0d 348 cr4 = read_cr4_safe();
9d975ebd
PE
349 printk("CR0: %08lx CR2: %08lx CR3: %08lx CR4: %08lx\n",
350 cr0, cr2, cr3, cr4);
bb1995d5
AS
351
352 get_debugreg(d0, 0);
353 get_debugreg(d1, 1);
354 get_debugreg(d2, 2);
355 get_debugreg(d3, 3);
356 printk("DR0: %08lx DR1: %08lx DR2: %08lx DR3: %08lx\n",
357 d0, d1, d2, d3);
9d975ebd 358
bb1995d5
AS
359 get_debugreg(d6, 6);
360 get_debugreg(d7, 7);
9d975ebd
PE
361 printk("DR6: %08lx DR7: %08lx\n",
362 d6, d7);
363}
bb1995d5 364
9d975ebd
PE
365void show_regs(struct pt_regs *regs)
366{
367 __show_registers(regs, 1);
5bc27dc2 368 show_trace(NULL, regs, &regs->sp, regs->bp);
1da177e4
LT
369}
370
371/*
65ea5b03
PA
372 * This gets run with %bx containing the
373 * function to call, and %dx containing
1da177e4
LT
374 * the "args".
375 */
376extern void kernel_thread_helper(void);
1da177e4
LT
377
378/*
379 * Create a kernel thread
380 */
381int kernel_thread(int (*fn)(void *), void * arg, unsigned long flags)
382{
383 struct pt_regs regs;
384
385 memset(&regs, 0, sizeof(regs));
386
65ea5b03
PA
387 regs.bx = (unsigned long) fn;
388 regs.dx = (unsigned long) arg;
1da177e4 389
65ea5b03
PA
390 regs.ds = __USER_DS;
391 regs.es = __USER_DS;
392 regs.fs = __KERNEL_PERCPU;
393 regs.orig_ax = -1;
394 regs.ip = (unsigned long) kernel_thread_helper;
395 regs.cs = __KERNEL_CS | get_kernel_rpl();
396 regs.flags = X86_EFLAGS_IF | X86_EFLAGS_SF | X86_EFLAGS_PF | 0x2;
1da177e4
LT
397
398 /* Ok, create the new process.. */
8cf2c519 399 return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, &regs, 0, NULL, NULL);
1da177e4 400}
129f6946 401EXPORT_SYMBOL(kernel_thread);
1da177e4
LT
402
403/*
404 * Free current thread data structures etc..
405 */
406void exit_thread(void)
407{
1da177e4 408 /* The process may have allocated an io port bitmap... nuke it. */
b3cf2576
SE
409 if (unlikely(test_thread_flag(TIF_IO_BITMAP))) {
410 struct task_struct *tsk = current;
411 struct thread_struct *t = &tsk->thread;
1da177e4
LT
412 int cpu = get_cpu();
413 struct tss_struct *tss = &per_cpu(init_tss, cpu);
414
415 kfree(t->io_bitmap_ptr);
416 t->io_bitmap_ptr = NULL;
b3cf2576 417 clear_thread_flag(TIF_IO_BITMAP);
1da177e4
LT
418 /*
419 * Careful, clear this in the TSS too:
420 */
421 memset(tss->io_bitmap, 0xff, tss->io_bitmap_max);
422 t->io_bitmap_max = 0;
423 tss->io_bitmap_owner = NULL;
424 tss->io_bitmap_max = 0;
a75c54f9 425 tss->x86_tss.io_bitmap_base = INVALID_IO_BITMAP_OFFSET;
1da177e4
LT
426 put_cpu();
427 }
428}
429
430void flush_thread(void)
431{
432 struct task_struct *tsk = current;
433
0f534093
RM
434 tsk->thread.debugreg0 = 0;
435 tsk->thread.debugreg1 = 0;
436 tsk->thread.debugreg2 = 0;
437 tsk->thread.debugreg3 = 0;
438 tsk->thread.debugreg6 = 0;
439 tsk->thread.debugreg7 = 0;
1da177e4 440 memset(tsk->thread.tls_array, 0, sizeof(tsk->thread.tls_array));
b3cf2576 441 clear_tsk_thread_flag(tsk, TIF_DEBUG);
1da177e4
LT
442 /*
443 * Forget coprocessor state..
444 */
445 clear_fpu(tsk);
446 clear_used_math();
447}
448
449void release_thread(struct task_struct *dead_task)
450{
2684927c 451 BUG_ON(dead_task->mm);
1da177e4
LT
452 release_vm86_irqs(dead_task);
453}
454
455/*
456 * This gets called before we allocate a new thread and copy
457 * the current task into it.
458 */
459void prepare_to_copy(struct task_struct *tsk)
460{
461 unlazy_fpu(tsk);
462}
463
65ea5b03 464int copy_thread(int nr, unsigned long clone_flags, unsigned long sp,
1da177e4
LT
465 unsigned long unused,
466 struct task_struct * p, struct pt_regs * regs)
467{
468 struct pt_regs * childregs;
469 struct task_struct *tsk;
470 int err;
471
07b047fc 472 childregs = task_pt_regs(p);
f48d9663 473 *childregs = *regs;
65ea5b03
PA
474 childregs->ax = 0;
475 childregs->sp = sp;
f48d9663 476
faca6227
PA
477 p->thread.sp = (unsigned long) childregs;
478 p->thread.sp0 = (unsigned long) (childregs+1);
1da177e4 479
faca6227 480 p->thread.ip = (unsigned long) ret_from_fork;
1da177e4 481
6612538c 482 savesegment(gs, p->thread.gs);
1da177e4
LT
483
484 tsk = current;
b3cf2576 485 if (unlikely(test_tsk_thread_flag(tsk, TIF_IO_BITMAP))) {
52978be6
AD
486 p->thread.io_bitmap_ptr = kmemdup(tsk->thread.io_bitmap_ptr,
487 IO_BITMAP_BYTES, GFP_KERNEL);
1da177e4
LT
488 if (!p->thread.io_bitmap_ptr) {
489 p->thread.io_bitmap_max = 0;
490 return -ENOMEM;
491 }
b3cf2576 492 set_tsk_thread_flag(p, TIF_IO_BITMAP);
1da177e4
LT
493 }
494
efd1ca52
RM
495 err = 0;
496
1da177e4
LT
497 /*
498 * Set a new TLS for the child thread?
499 */
efd1ca52
RM
500 if (clone_flags & CLONE_SETTLS)
501 err = do_set_thread_area(p, -1,
65ea5b03 502 (struct user_desc __user *)childregs->si, 0);
1da177e4 503
1da177e4
LT
504 if (err && p->thread.io_bitmap_ptr) {
505 kfree(p->thread.io_bitmap_ptr);
506 p->thread.io_bitmap_max = 0;
507 }
508 return err;
509}
510
513ad84b
IM
511void
512start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp)
513{
514 __asm__("movl %0, %%gs" :: "r"(0));
515 regs->fs = 0;
516 set_fs(USER_DS);
517 regs->ds = __USER_DS;
518 regs->es = __USER_DS;
519 regs->ss = __USER_DS;
520 regs->cs = __USER_CS;
521 regs->ip = new_ip;
522 regs->sp = new_sp;
523}
524EXPORT_SYMBOL_GPL(start_thread);
525
cf99abac 526#ifdef CONFIG_SECCOMP
bdb4f156 527static void hard_disable_TSC(void)
cf99abac
AA
528{
529 write_cr4(read_cr4() | X86_CR4_TSD);
530}
531void disable_TSC(void)
532{
533 preempt_disable();
534 if (!test_and_set_thread_flag(TIF_NOTSC))
535 /*
536 * Must flip the CPU state synchronously with
537 * TIF_NOTSC in the current running context.
538 */
539 hard_disable_TSC();
540 preempt_enable();
541}
bdb4f156 542static void hard_enable_TSC(void)
cf99abac
AA
543{
544 write_cr4(read_cr4() & ~X86_CR4_TSD);
545}
546#endif /* CONFIG_SECCOMP */
547
548static noinline void
549__switch_to_xtra(struct task_struct *prev_p, struct task_struct *next_p,
550 struct tss_struct *tss)
1da177e4 551{
7e991604 552 struct thread_struct *prev, *next;
eee3af4a 553 unsigned long debugctl;
b3cf2576 554
7e991604 555 prev = &prev_p->thread;
b3cf2576
SE
556 next = &next_p->thread;
557
eee3af4a
MM
558 debugctl = prev->debugctlmsr;
559 if (next->ds_area_msr != prev->ds_area_msr) {
560 /* we clear debugctl to make sure DS
561 * is not in use when we change it */
562 debugctl = 0;
5b0e5084 563 update_debugctlmsr(0);
eee3af4a
MM
564 wrmsr(MSR_IA32_DS_AREA, next->ds_area_msr, 0);
565 }
566
567 if (next->debugctlmsr != debugctl)
5b0e5084 568 update_debugctlmsr(next->debugctlmsr);
7e991604 569
b3cf2576 570 if (test_tsk_thread_flag(next_p, TIF_DEBUG)) {
0f534093
RM
571 set_debugreg(next->debugreg0, 0);
572 set_debugreg(next->debugreg1, 1);
573 set_debugreg(next->debugreg2, 2);
574 set_debugreg(next->debugreg3, 3);
b3cf2576 575 /* no 4 and 5 */
0f534093
RM
576 set_debugreg(next->debugreg6, 6);
577 set_debugreg(next->debugreg7, 7);
b3cf2576
SE
578 }
579
cf99abac
AA
580#ifdef CONFIG_SECCOMP
581 if (test_tsk_thread_flag(prev_p, TIF_NOTSC) ^
582 test_tsk_thread_flag(next_p, TIF_NOTSC)) {
583 /* prev and next are different */
584 if (test_tsk_thread_flag(next_p, TIF_NOTSC))
585 hard_disable_TSC();
586 else
587 hard_enable_TSC();
588 }
589#endif
590
b4ef95de 591#ifdef X86_BTS
eee3af4a
MM
592 if (test_tsk_thread_flag(prev_p, TIF_BTS_TRACE_TS))
593 ptrace_bts_take_timestamp(prev_p, BTS_TASK_DEPARTS);
594
595 if (test_tsk_thread_flag(next_p, TIF_BTS_TRACE_TS))
596 ptrace_bts_take_timestamp(next_p, BTS_TASK_ARRIVES);
b4ef95de 597#endif
eee3af4a
MM
598
599
b3cf2576 600 if (!test_tsk_thread_flag(next_p, TIF_IO_BITMAP)) {
1da177e4
LT
601 /*
602 * Disable the bitmap via an invalid offset. We still cache
603 * the previous bitmap owner and the IO bitmap contents:
604 */
a75c54f9 605 tss->x86_tss.io_bitmap_base = INVALID_IO_BITMAP_OFFSET;
1da177e4
LT
606 return;
607 }
b3cf2576 608
1da177e4
LT
609 if (likely(next == tss->io_bitmap_owner)) {
610 /*
611 * Previous owner of the bitmap (hence the bitmap content)
612 * matches the next task, we dont have to do anything but
613 * to set a valid offset in the TSS:
614 */
a75c54f9 615 tss->x86_tss.io_bitmap_base = IO_BITMAP_OFFSET;
1da177e4
LT
616 return;
617 }
618 /*
619 * Lazy TSS's I/O bitmap copy. We set an invalid offset here
620 * and we let the task to get a GPF in case an I/O instruction
621 * is performed. The handler of the GPF will verify that the
622 * faulting task has a valid I/O bitmap and, it true, does the
623 * real copy and restart the instruction. This will save us
624 * redundant copies when the currently switched task does not
625 * perform any I/O during its timeslice.
626 */
a75c54f9 627 tss->x86_tss.io_bitmap_base = INVALID_IO_BITMAP_OFFSET_LAZY;
1da177e4 628}
1da177e4
LT
629
630/*
631 * switch_to(x,yn) should switch tasks from x to y.
632 *
633 * We fsave/fwait so that an exception goes off at the right time
634 * (as a call from the fsave or fwait in effect) rather than to
635 * the wrong process. Lazy FP saving no longer makes any sense
636 * with modern CPU's, and this simplifies a lot of things (SMP
637 * and UP become the same).
638 *
639 * NOTE! We used to use the x86 hardware context switching. The
640 * reason for not using it any more becomes apparent when you
641 * try to recover gracefully from saved state that is no longer
642 * valid (stale segment register values in particular). With the
643 * hardware task-switch, there is no way to fix up bad state in
644 * a reasonable manner.
645 *
646 * The fact that Intel documents the hardware task-switching to
647 * be slow is a fairly red herring - this code is not noticeably
648 * faster. However, there _is_ some room for improvement here,
649 * so the performance issues may eventually be a valid point.
650 * More important, however, is the fact that this allows us much
651 * more flexibility.
652 *
65ea5b03 653 * The return value (in %ax) will be the "prev" task after
1da177e4
LT
654 * the task-switch, and shows up in ret_from_fork in entry.S,
655 * for example.
656 */
75604d7f 657struct task_struct * __switch_to(struct task_struct *prev_p, struct task_struct *next_p)
1da177e4
LT
658{
659 struct thread_struct *prev = &prev_p->thread,
660 *next = &next_p->thread;
661 int cpu = smp_processor_id();
662 struct tss_struct *tss = &per_cpu(init_tss, cpu);
663
664 /* never put a printk in __switch_to... printk() calls wake_up*() indirectly */
665
666 __unlazy_fpu(prev_p);
667
acc20761
CE
668
669 /* we're going to use this soon, after a few expensive things */
670 if (next_p->fpu_counter > 5)
671 prefetch(&next->i387.fxsave);
672
1da177e4 673 /*
e7a2ff59 674 * Reload esp0.
1da177e4 675 */
faca6227 676 load_sp0(tss, next);
1da177e4
LT
677
678 /*
464d1a78 679 * Save away %gs. No need to save %fs, as it was saved on the
f95d47ca
JF
680 * stack on entry. No need to save %es and %ds, as those are
681 * always kernel segments while inside the kernel. Doing this
682 * before setting the new TLS descriptors avoids the situation
683 * where we temporarily have non-reloadable segments in %fs
684 * and %gs. This could be an issue if the NMI handler ever
685 * used %fs or %gs (it does not today), or if the kernel is
686 * running inside of a hypervisor layer.
1da177e4 687 */
464d1a78 688 savesegment(gs, prev->gs);
1da177e4
LT
689
690 /*
e7a2ff59 691 * Load the per-thread Thread-Local Storage descriptor.
1da177e4 692 */
e7a2ff59 693 load_TLS(next, cpu);
1da177e4 694
8b151144
ZA
695 /*
696 * Restore IOPL if needed. In normal use, the flags restore
697 * in the switch assembly will handle this. But if the kernel
698 * is running virtualized at a non-zero CPL, the popf will
699 * not restore flags, so it must be done in a separate step.
700 */
701 if (get_kernel_rpl() && unlikely(prev->iopl != next->iopl))
702 set_iopl_mask(next->iopl);
703
1da177e4 704 /*
b3cf2576 705 * Now maybe handle debug registers and/or IO bitmaps
1da177e4 706 */
cf99abac
AA
707 if (unlikely(task_thread_info(prev_p)->flags & _TIF_WORK_CTXSW_PREV ||
708 task_thread_info(next_p)->flags & _TIF_WORK_CTXSW_NEXT))
709 __switch_to_xtra(prev_p, next_p, tss);
ffaa8bd6 710
9226d125
ZA
711 /*
712 * Leave lazy mode, flushing any hypercalls made here.
713 * This must be done before restoring TLS segments so
714 * the GDT and LDT are properly updated, and must be
715 * done before math_state_restore, so the TS bit is up
716 * to date.
717 */
718 arch_leave_lazy_cpu_mode();
719
acc20761
CE
720 /* If the task has used fpu the last 5 timeslices, just do a full
721 * restore of the math state immediately to avoid the trap; the
722 * chances of needing FPU soon are obviously high now
723 */
724 if (next_p->fpu_counter > 5)
725 math_state_restore();
726
9226d125
ZA
727 /*
728 * Restore %gs if needed (which is common)
729 */
730 if (prev->gs | next->gs)
731 loadsegment(gs, next->gs);
732
7c3576d2 733 x86_write_percpu(current_task, next_p);
9226d125 734
1da177e4
LT
735 return prev_p;
736}
737
738asmlinkage int sys_fork(struct pt_regs regs)
739{
65ea5b03 740 return do_fork(SIGCHLD, regs.sp, &regs, 0, NULL, NULL);
1da177e4
LT
741}
742
743asmlinkage int sys_clone(struct pt_regs regs)
744{
745 unsigned long clone_flags;
746 unsigned long newsp;
747 int __user *parent_tidptr, *child_tidptr;
748
65ea5b03
PA
749 clone_flags = regs.bx;
750 newsp = regs.cx;
751 parent_tidptr = (int __user *)regs.dx;
752 child_tidptr = (int __user *)regs.di;
1da177e4 753 if (!newsp)
65ea5b03 754 newsp = regs.sp;
1da177e4
LT
755 return do_fork(clone_flags, newsp, &regs, 0, parent_tidptr, child_tidptr);
756}
757
758/*
759 * This is trivial, and on the face of it looks like it
760 * could equally well be done in user mode.
761 *
762 * Not so, for quite unobvious reasons - register pressure.
763 * In user mode vfork() cannot have a stack frame, and if
764 * done by calling the "clone()" system call directly, you
765 * do not have enough call-clobbered registers to hold all
766 * the information you need.
767 */
768asmlinkage int sys_vfork(struct pt_regs regs)
769{
65ea5b03 770 return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, regs.sp, &regs, 0, NULL, NULL);
1da177e4
LT
771}
772
773/*
774 * sys_execve() executes a new program.
775 */
776asmlinkage int sys_execve(struct pt_regs regs)
777{
778 int error;
779 char * filename;
780
65ea5b03 781 filename = getname((char __user *) regs.bx);
1da177e4
LT
782 error = PTR_ERR(filename);
783 if (IS_ERR(filename))
784 goto out;
785 error = do_execve(filename,
65ea5b03
PA
786 (char __user * __user *) regs.cx,
787 (char __user * __user *) regs.dx,
1da177e4
LT
788 &regs);
789 if (error == 0) {
1da177e4
LT
790 /* Make sure we don't return using sysenter.. */
791 set_thread_flag(TIF_IRET);
792 }
793 putname(filename);
794out:
795 return error;
796}
797
798#define top_esp (THREAD_SIZE - sizeof(unsigned long))
799#define top_ebp (THREAD_SIZE - 2*sizeof(unsigned long))
800
801unsigned long get_wchan(struct task_struct *p)
802{
65ea5b03 803 unsigned long bp, sp, ip;
1da177e4
LT
804 unsigned long stack_page;
805 int count = 0;
806 if (!p || p == current || p->state == TASK_RUNNING)
807 return 0;
65e0fdff 808 stack_page = (unsigned long)task_stack_page(p);
faca6227 809 sp = p->thread.sp;
65ea5b03 810 if (!stack_page || sp < stack_page || sp > top_esp+stack_page)
1da177e4 811 return 0;
65ea5b03
PA
812 /* include/asm-i386/system.h:switch_to() pushes bp last. */
813 bp = *(unsigned long *) sp;
1da177e4 814 do {
65ea5b03 815 if (bp < stack_page || bp > top_ebp+stack_page)
1da177e4 816 return 0;
65ea5b03
PA
817 ip = *(unsigned long *) (bp+4);
818 if (!in_sched_functions(ip))
819 return ip;
820 bp = *(unsigned long *) bp;
1da177e4
LT
821 } while (count++ < 16);
822 return 0;
823}
824
1da177e4
LT
825unsigned long arch_align_stack(unsigned long sp)
826{
c16b63e0 827 if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space)
1da177e4
LT
828 sp -= get_random_int() % 8192;
829 return sp & ~0xf;
830}
c1d171a0
JK
831
832unsigned long arch_randomize_brk(struct mm_struct *mm)
833{
834 unsigned long range_end = mm->brk + 0x02000000;
835 return randomize_range(mm->brk, range_end, 0) ? : mm->brk;
836}