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> |
90f7d25c | 40 | #include <linux/dmi.h> |
8b96f011 | 41 | #include <linux/ftrace.h> |
1da177e4 LT |
42 | |
43 | #include <asm/uaccess.h> | |
44 | #include <asm/pgtable.h> | |
45 | #include <asm/system.h> | |
46 | #include <asm/io.h> | |
47 | #include <asm/ldt.h> | |
48 | #include <asm/processor.h> | |
49 | #include <asm/i387.h> | |
1da177e4 LT |
50 | #include <asm/desc.h> |
51 | #ifdef CONFIG_MATH_EMULATION | |
52 | #include <asm/math_emu.h> | |
53 | #endif | |
54 | ||
1da177e4 LT |
55 | #include <linux/err.h> |
56 | ||
f3705136 ZM |
57 | #include <asm/tlbflush.h> |
58 | #include <asm/cpu.h> | |
718fc13b | 59 | #include <asm/kdebug.h> |
1eda8149 | 60 | #include <asm/idle.h> |
bbc1f698 | 61 | #include <asm/syscalls.h> |
fb26132b | 62 | #include <asm/smp.h> |
f3705136 | 63 | |
1da177e4 LT |
64 | asmlinkage void ret_from_fork(void) __asm__("ret_from_fork"); |
65 | ||
7c3576d2 JF |
66 | DEFINE_PER_CPU(struct task_struct *, current_task) = &init_task; |
67 | EXPORT_PER_CPU_SYMBOL(current_task); | |
68 | ||
69 | DEFINE_PER_CPU(int, cpu_number); | |
70 | EXPORT_PER_CPU_SYMBOL(cpu_number); | |
71 | ||
1da177e4 LT |
72 | /* |
73 | * Return saved PC of a blocked thread. | |
74 | */ | |
75 | unsigned long thread_saved_pc(struct task_struct *tsk) | |
76 | { | |
faca6227 | 77 | return ((unsigned long *)tsk->thread.sp)[3]; |
1da177e4 LT |
78 | } |
79 | ||
913da64b AN |
80 | #ifndef CONFIG_SMP |
81 | static inline void play_dead(void) | |
82 | { | |
83 | BUG(); | |
84 | } | |
85 | #endif | |
86 | ||
1da177e4 LT |
87 | /* |
88 | * The idle thread. There's no useful work to be | |
89 | * done, so just try to conserve power and have a | |
90 | * low exit latency (ie sit in a loop waiting for | |
91 | * somebody to say that they'd like to reschedule) | |
92 | */ | |
f3705136 | 93 | void cpu_idle(void) |
1da177e4 | 94 | { |
5bfb5d69 | 95 | int cpu = smp_processor_id(); |
f3705136 | 96 | |
495ab9c0 | 97 | current_thread_info()->status |= TS_POLLING; |
64c7c8f8 | 98 | |
1da177e4 LT |
99 | /* endless idle loop with no priority at all */ |
100 | while (1) { | |
b8f8c3cf | 101 | tick_nohz_stop_sched_tick(1); |
1da177e4 | 102 | while (!need_resched()) { |
1da177e4 | 103 | |
f1d1a842 | 104 | check_pgt_cache(); |
1da177e4 | 105 | rmb(); |
1da177e4 | 106 | |
0723a69a BL |
107 | if (rcu_pending(cpu)) |
108 | rcu_check_callbacks(cpu, 0); | |
109 | ||
f3705136 ZM |
110 | if (cpu_is_offline(cpu)) |
111 | play_dead(); | |
112 | ||
7f424a8b | 113 | local_irq_disable(); |
1da177e4 | 114 | __get_cpu_var(irq_stat).idle_timestamp = jiffies; |
6cd8a4bb SR |
115 | /* Don't trace irqs off for idle */ |
116 | stop_critical_timings(); | |
6ddd2a27 | 117 | pm_idle(); |
6cd8a4bb | 118 | start_critical_timings(); |
1da177e4 | 119 | } |
74167347 | 120 | tick_nohz_restart_sched_tick(); |
5bfb5d69 | 121 | preempt_enable_no_resched(); |
1da177e4 | 122 | schedule(); |
5bfb5d69 | 123 | preempt_disable(); |
1da177e4 LT |
124 | } |
125 | } | |
126 | ||
e2ce07c8 | 127 | void __show_regs(struct pt_regs *regs, int all) |
1da177e4 LT |
128 | { |
129 | unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L; | |
bb1995d5 | 130 | unsigned long d0, d1, d2, d3, d6, d7; |
65ea5b03 | 131 | unsigned long sp; |
9d975ebd | 132 | unsigned short ss, gs; |
90f7d25c | 133 | const char *board; |
9d975ebd PE |
134 | |
135 | if (user_mode_vm(regs)) { | |
65ea5b03 PA |
136 | sp = regs->sp; |
137 | ss = regs->ss & 0xffff; | |
9d975ebd PE |
138 | savesegment(gs, gs); |
139 | } else { | |
65ea5b03 | 140 | sp = (unsigned long) (®s->sp); |
9d975ebd PE |
141 | savesegment(ss, ss); |
142 | savesegment(gs, gs); | |
143 | } | |
1da177e4 LT |
144 | |
145 | printk("\n"); | |
90f7d25c AV |
146 | |
147 | board = dmi_get_system_info(DMI_PRODUCT_NAME); | |
148 | if (!board) | |
149 | board = ""; | |
150 | printk("Pid: %d, comm: %s %s (%s %.*s) %s\n", | |
60812a4a | 151 | task_pid_nr(current), current->comm, |
9d975ebd PE |
152 | print_tainted(), init_utsname()->release, |
153 | (int)strcspn(init_utsname()->version, " "), | |
90f7d25c | 154 | init_utsname()->version, board); |
9d975ebd PE |
155 | |
156 | printk("EIP: %04x:[<%08lx>] EFLAGS: %08lx CPU: %d\n", | |
92bc2056 | 157 | (u16)regs->cs, regs->ip, regs->flags, |
9d975ebd | 158 | smp_processor_id()); |
65ea5b03 | 159 | print_symbol("EIP is at %s\n", regs->ip); |
1da177e4 | 160 | |
1da177e4 | 161 | printk("EAX: %08lx EBX: %08lx ECX: %08lx EDX: %08lx\n", |
65ea5b03 | 162 | regs->ax, regs->bx, regs->cx, regs->dx); |
9d975ebd | 163 | printk("ESI: %08lx EDI: %08lx EBP: %08lx ESP: %08lx\n", |
65ea5b03 | 164 | regs->si, regs->di, regs->bp, sp); |
9d975ebd | 165 | printk(" DS: %04x ES: %04x FS: %04x GS: %04x SS: %04x\n", |
92bc2056 | 166 | (u16)regs->ds, (u16)regs->es, (u16)regs->fs, gs, ss); |
9d975ebd PE |
167 | |
168 | if (!all) | |
169 | return; | |
1da177e4 | 170 | |
4bb0d3ec ZA |
171 | cr0 = read_cr0(); |
172 | cr2 = read_cr2(); | |
173 | cr3 = read_cr3(); | |
ff6e8c0d | 174 | cr4 = read_cr4_safe(); |
9d975ebd PE |
175 | printk("CR0: %08lx CR2: %08lx CR3: %08lx CR4: %08lx\n", |
176 | cr0, cr2, cr3, cr4); | |
bb1995d5 AS |
177 | |
178 | get_debugreg(d0, 0); | |
179 | get_debugreg(d1, 1); | |
180 | get_debugreg(d2, 2); | |
181 | get_debugreg(d3, 3); | |
182 | printk("DR0: %08lx DR1: %08lx DR2: %08lx DR3: %08lx\n", | |
183 | d0, d1, d2, d3); | |
9d975ebd | 184 | |
bb1995d5 AS |
185 | get_debugreg(d6, 6); |
186 | get_debugreg(d7, 7); | |
9d975ebd PE |
187 | printk("DR6: %08lx DR7: %08lx\n", |
188 | d6, d7); | |
189 | } | |
bb1995d5 | 190 | |
9d975ebd PE |
191 | void show_regs(struct pt_regs *regs) |
192 | { | |
e2ce07c8 | 193 | __show_regs(regs, 1); |
5bc27dc2 | 194 | show_trace(NULL, regs, ®s->sp, regs->bp); |
1da177e4 LT |
195 | } |
196 | ||
197 | /* | |
65ea5b03 PA |
198 | * This gets run with %bx containing the |
199 | * function to call, and %dx containing | |
1da177e4 LT |
200 | * the "args". |
201 | */ | |
202 | extern void kernel_thread_helper(void); | |
1da177e4 LT |
203 | |
204 | /* | |
205 | * Create a kernel thread | |
206 | */ | |
207 | int kernel_thread(int (*fn)(void *), void * arg, unsigned long flags) | |
208 | { | |
209 | struct pt_regs regs; | |
210 | ||
211 | memset(®s, 0, sizeof(regs)); | |
212 | ||
65ea5b03 PA |
213 | regs.bx = (unsigned long) fn; |
214 | regs.dx = (unsigned long) arg; | |
1da177e4 | 215 | |
65ea5b03 PA |
216 | regs.ds = __USER_DS; |
217 | regs.es = __USER_DS; | |
218 | regs.fs = __KERNEL_PERCPU; | |
219 | regs.orig_ax = -1; | |
220 | regs.ip = (unsigned long) kernel_thread_helper; | |
221 | regs.cs = __KERNEL_CS | get_kernel_rpl(); | |
222 | regs.flags = X86_EFLAGS_IF | X86_EFLAGS_SF | X86_EFLAGS_PF | 0x2; | |
1da177e4 LT |
223 | |
224 | /* Ok, create the new process.. */ | |
8cf2c519 | 225 | return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, ®s, 0, NULL, NULL); |
1da177e4 | 226 | } |
129f6946 | 227 | EXPORT_SYMBOL(kernel_thread); |
1da177e4 LT |
228 | |
229 | /* | |
230 | * Free current thread data structures etc.. | |
231 | */ | |
232 | void exit_thread(void) | |
233 | { | |
1da177e4 | 234 | /* The process may have allocated an io port bitmap... nuke it. */ |
b3cf2576 SE |
235 | if (unlikely(test_thread_flag(TIF_IO_BITMAP))) { |
236 | struct task_struct *tsk = current; | |
237 | struct thread_struct *t = &tsk->thread; | |
1da177e4 LT |
238 | int cpu = get_cpu(); |
239 | struct tss_struct *tss = &per_cpu(init_tss, cpu); | |
240 | ||
241 | kfree(t->io_bitmap_ptr); | |
242 | t->io_bitmap_ptr = NULL; | |
b3cf2576 | 243 | clear_thread_flag(TIF_IO_BITMAP); |
1da177e4 LT |
244 | /* |
245 | * Careful, clear this in the TSS too: | |
246 | */ | |
247 | memset(tss->io_bitmap, 0xff, tss->io_bitmap_max); | |
248 | t->io_bitmap_max = 0; | |
249 | tss->io_bitmap_owner = NULL; | |
250 | tss->io_bitmap_max = 0; | |
a75c54f9 | 251 | tss->x86_tss.io_bitmap_base = INVALID_IO_BITMAP_OFFSET; |
1da177e4 LT |
252 | put_cpu(); |
253 | } | |
93fa7636 MM |
254 | #ifdef CONFIG_X86_DS |
255 | /* Free any DS contexts that have not been properly released. */ | |
256 | if (unlikely(current->thread.ds_ctx)) { | |
257 | /* we clear debugctl to make sure DS is not used. */ | |
258 | update_debugctlmsr(0); | |
259 | ds_free(current->thread.ds_ctx); | |
260 | } | |
261 | #endif /* CONFIG_X86_DS */ | |
1da177e4 LT |
262 | } |
263 | ||
264 | void flush_thread(void) | |
265 | { | |
266 | struct task_struct *tsk = current; | |
267 | ||
0f534093 RM |
268 | tsk->thread.debugreg0 = 0; |
269 | tsk->thread.debugreg1 = 0; | |
270 | tsk->thread.debugreg2 = 0; | |
271 | tsk->thread.debugreg3 = 0; | |
272 | tsk->thread.debugreg6 = 0; | |
273 | tsk->thread.debugreg7 = 0; | |
1da177e4 | 274 | memset(tsk->thread.tls_array, 0, sizeof(tsk->thread.tls_array)); |
b3cf2576 | 275 | clear_tsk_thread_flag(tsk, TIF_DEBUG); |
1da177e4 LT |
276 | /* |
277 | * Forget coprocessor state.. | |
278 | */ | |
75118a82 | 279 | tsk->fpu_counter = 0; |
1da177e4 LT |
280 | clear_fpu(tsk); |
281 | clear_used_math(); | |
282 | } | |
283 | ||
284 | void release_thread(struct task_struct *dead_task) | |
285 | { | |
2684927c | 286 | BUG_ON(dead_task->mm); |
1da177e4 LT |
287 | release_vm86_irqs(dead_task); |
288 | } | |
289 | ||
290 | /* | |
291 | * This gets called before we allocate a new thread and copy | |
292 | * the current task into it. | |
293 | */ | |
294 | void prepare_to_copy(struct task_struct *tsk) | |
295 | { | |
296 | unlazy_fpu(tsk); | |
297 | } | |
298 | ||
65ea5b03 | 299 | int copy_thread(int nr, unsigned long clone_flags, unsigned long sp, |
1da177e4 LT |
300 | unsigned long unused, |
301 | struct task_struct * p, struct pt_regs * regs) | |
302 | { | |
303 | struct pt_regs * childregs; | |
304 | struct task_struct *tsk; | |
305 | int err; | |
306 | ||
07b047fc | 307 | childregs = task_pt_regs(p); |
f48d9663 | 308 | *childregs = *regs; |
65ea5b03 PA |
309 | childregs->ax = 0; |
310 | childregs->sp = sp; | |
f48d9663 | 311 | |
faca6227 PA |
312 | p->thread.sp = (unsigned long) childregs; |
313 | p->thread.sp0 = (unsigned long) (childregs+1); | |
1da177e4 | 314 | |
faca6227 | 315 | p->thread.ip = (unsigned long) ret_from_fork; |
1da177e4 | 316 | |
6612538c | 317 | savesegment(gs, p->thread.gs); |
1da177e4 LT |
318 | |
319 | tsk = current; | |
b3cf2576 | 320 | if (unlikely(test_tsk_thread_flag(tsk, TIF_IO_BITMAP))) { |
52978be6 AD |
321 | p->thread.io_bitmap_ptr = kmemdup(tsk->thread.io_bitmap_ptr, |
322 | IO_BITMAP_BYTES, GFP_KERNEL); | |
1da177e4 LT |
323 | if (!p->thread.io_bitmap_ptr) { |
324 | p->thread.io_bitmap_max = 0; | |
325 | return -ENOMEM; | |
326 | } | |
b3cf2576 | 327 | set_tsk_thread_flag(p, TIF_IO_BITMAP); |
1da177e4 LT |
328 | } |
329 | ||
efd1ca52 RM |
330 | err = 0; |
331 | ||
1da177e4 LT |
332 | /* |
333 | * Set a new TLS for the child thread? | |
334 | */ | |
efd1ca52 RM |
335 | if (clone_flags & CLONE_SETTLS) |
336 | err = do_set_thread_area(p, -1, | |
65ea5b03 | 337 | (struct user_desc __user *)childregs->si, 0); |
1da177e4 | 338 | |
1da177e4 LT |
339 | if (err && p->thread.io_bitmap_ptr) { |
340 | kfree(p->thread.io_bitmap_ptr); | |
341 | p->thread.io_bitmap_max = 0; | |
342 | } | |
343 | return err; | |
344 | } | |
345 | ||
513ad84b IM |
346 | void |
347 | start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp) | |
348 | { | |
349 | __asm__("movl %0, %%gs" :: "r"(0)); | |
350 | regs->fs = 0; | |
351 | set_fs(USER_DS); | |
352 | regs->ds = __USER_DS; | |
353 | regs->es = __USER_DS; | |
354 | regs->ss = __USER_DS; | |
355 | regs->cs = __USER_CS; | |
356 | regs->ip = new_ip; | |
357 | regs->sp = new_sp; | |
aa283f49 SS |
358 | /* |
359 | * Free the old FP and other extended state | |
360 | */ | |
361 | free_thread_xstate(current); | |
513ad84b IM |
362 | } |
363 | EXPORT_SYMBOL_GPL(start_thread); | |
364 | ||
bdb4f156 | 365 | static void hard_disable_TSC(void) |
cf99abac AA |
366 | { |
367 | write_cr4(read_cr4() | X86_CR4_TSD); | |
368 | } | |
529e25f6 | 369 | |
cf99abac AA |
370 | void disable_TSC(void) |
371 | { | |
372 | preempt_disable(); | |
373 | if (!test_and_set_thread_flag(TIF_NOTSC)) | |
374 | /* | |
375 | * Must flip the CPU state synchronously with | |
376 | * TIF_NOTSC in the current running context. | |
377 | */ | |
378 | hard_disable_TSC(); | |
379 | preempt_enable(); | |
380 | } | |
529e25f6 | 381 | |
bdb4f156 | 382 | static void hard_enable_TSC(void) |
cf99abac AA |
383 | { |
384 | write_cr4(read_cr4() & ~X86_CR4_TSD); | |
385 | } | |
529e25f6 | 386 | |
a4928cff | 387 | static void enable_TSC(void) |
529e25f6 EB |
388 | { |
389 | preempt_disable(); | |
390 | if (test_and_clear_thread_flag(TIF_NOTSC)) | |
391 | /* | |
392 | * Must flip the CPU state synchronously with | |
393 | * TIF_NOTSC in the current running context. | |
394 | */ | |
395 | hard_enable_TSC(); | |
396 | preempt_enable(); | |
397 | } | |
398 | ||
399 | int get_tsc_mode(unsigned long adr) | |
400 | { | |
401 | unsigned int val; | |
402 | ||
403 | if (test_thread_flag(TIF_NOTSC)) | |
404 | val = PR_TSC_SIGSEGV; | |
405 | else | |
406 | val = PR_TSC_ENABLE; | |
407 | ||
408 | return put_user(val, (unsigned int __user *)adr); | |
409 | } | |
410 | ||
411 | int set_tsc_mode(unsigned int val) | |
412 | { | |
413 | if (val == PR_TSC_SIGSEGV) | |
414 | disable_TSC(); | |
415 | else if (val == PR_TSC_ENABLE) | |
416 | enable_TSC(); | |
417 | else | |
418 | return -EINVAL; | |
419 | ||
420 | return 0; | |
421 | } | |
cf99abac | 422 | |
970e7250 AM |
423 | #ifdef CONFIG_X86_DS |
424 | static int update_debugctl(struct thread_struct *prev, | |
425 | struct thread_struct *next, unsigned long debugctl) | |
1da177e4 | 426 | { |
970e7250 AM |
427 | unsigned long ds_prev = 0; |
428 | unsigned long ds_next = 0; | |
b3cf2576 | 429 | |
93fa7636 MM |
430 | if (prev->ds_ctx) |
431 | ds_prev = (unsigned long)prev->ds_ctx->ds; | |
432 | if (next->ds_ctx) | |
433 | ds_next = (unsigned long)next->ds_ctx->ds; | |
434 | ||
435 | if (ds_next != ds_prev) { | |
eee3af4a MM |
436 | /* we clear debugctl to make sure DS |
437 | * is not in use when we change it */ | |
438 | debugctl = 0; | |
5b0e5084 | 439 | update_debugctlmsr(0); |
93fa7636 | 440 | wrmsr(MSR_IA32_DS_AREA, ds_next, 0); |
eee3af4a | 441 | } |
970e7250 AM |
442 | return debugctl; |
443 | } | |
444 | #else | |
445 | static int update_debugctl(struct thread_struct *prev, | |
446 | struct thread_struct *next, unsigned long debugctl) | |
447 | { | |
448 | return debugctl; | |
449 | } | |
93fa7636 | 450 | #endif /* CONFIG_X86_DS */ |
eee3af4a | 451 | |
cf99abac AA |
452 | static noinline void |
453 | __switch_to_xtra(struct task_struct *prev_p, struct task_struct *next_p, | |
454 | struct tss_struct *tss) | |
1da177e4 | 455 | { |
7e991604 | 456 | struct thread_struct *prev, *next; |
eee3af4a | 457 | unsigned long debugctl; |
b3cf2576 | 458 | |
7e991604 | 459 | prev = &prev_p->thread; |
b3cf2576 SE |
460 | next = &next_p->thread; |
461 | ||
970e7250 | 462 | debugctl = update_debugctl(prev, next, prev->debugctlmsr); |
eee3af4a MM |
463 | |
464 | if (next->debugctlmsr != debugctl) | |
5b0e5084 | 465 | update_debugctlmsr(next->debugctlmsr); |
7e991604 | 466 | |
b3cf2576 | 467 | if (test_tsk_thread_flag(next_p, TIF_DEBUG)) { |
0f534093 RM |
468 | set_debugreg(next->debugreg0, 0); |
469 | set_debugreg(next->debugreg1, 1); | |
470 | set_debugreg(next->debugreg2, 2); | |
471 | set_debugreg(next->debugreg3, 3); | |
b3cf2576 | 472 | /* no 4 and 5 */ |
0f534093 RM |
473 | set_debugreg(next->debugreg6, 6); |
474 | set_debugreg(next->debugreg7, 7); | |
b3cf2576 SE |
475 | } |
476 | ||
cf99abac AA |
477 | if (test_tsk_thread_flag(prev_p, TIF_NOTSC) ^ |
478 | test_tsk_thread_flag(next_p, TIF_NOTSC)) { | |
479 | /* prev and next are different */ | |
480 | if (test_tsk_thread_flag(next_p, TIF_NOTSC)) | |
481 | hard_disable_TSC(); | |
482 | else | |
483 | hard_enable_TSC(); | |
484 | } | |
cf99abac | 485 | |
93fa7636 | 486 | #ifdef CONFIG_X86_PTRACE_BTS |
eee3af4a MM |
487 | if (test_tsk_thread_flag(prev_p, TIF_BTS_TRACE_TS)) |
488 | ptrace_bts_take_timestamp(prev_p, BTS_TASK_DEPARTS); | |
489 | ||
490 | if (test_tsk_thread_flag(next_p, TIF_BTS_TRACE_TS)) | |
491 | ptrace_bts_take_timestamp(next_p, BTS_TASK_ARRIVES); | |
93fa7636 | 492 | #endif /* CONFIG_X86_PTRACE_BTS */ |
eee3af4a MM |
493 | |
494 | ||
b3cf2576 | 495 | if (!test_tsk_thread_flag(next_p, TIF_IO_BITMAP)) { |
1da177e4 LT |
496 | /* |
497 | * Disable the bitmap via an invalid offset. We still cache | |
498 | * the previous bitmap owner and the IO bitmap contents: | |
499 | */ | |
a75c54f9 | 500 | tss->x86_tss.io_bitmap_base = INVALID_IO_BITMAP_OFFSET; |
1da177e4 LT |
501 | return; |
502 | } | |
b3cf2576 | 503 | |
1da177e4 LT |
504 | if (likely(next == tss->io_bitmap_owner)) { |
505 | /* | |
506 | * Previous owner of the bitmap (hence the bitmap content) | |
507 | * matches the next task, we dont have to do anything but | |
508 | * to set a valid offset in the TSS: | |
509 | */ | |
a75c54f9 | 510 | tss->x86_tss.io_bitmap_base = IO_BITMAP_OFFSET; |
1da177e4 LT |
511 | return; |
512 | } | |
513 | /* | |
514 | * Lazy TSS's I/O bitmap copy. We set an invalid offset here | |
515 | * and we let the task to get a GPF in case an I/O instruction | |
516 | * is performed. The handler of the GPF will verify that the | |
517 | * faulting task has a valid I/O bitmap and, it true, does the | |
518 | * real copy and restart the instruction. This will save us | |
519 | * redundant copies when the currently switched task does not | |
520 | * perform any I/O during its timeslice. | |
521 | */ | |
a75c54f9 | 522 | tss->x86_tss.io_bitmap_base = INVALID_IO_BITMAP_OFFSET_LAZY; |
1da177e4 | 523 | } |
1da177e4 LT |
524 | |
525 | /* | |
526 | * switch_to(x,yn) should switch tasks from x to y. | |
527 | * | |
528 | * We fsave/fwait so that an exception goes off at the right time | |
529 | * (as a call from the fsave or fwait in effect) rather than to | |
530 | * the wrong process. Lazy FP saving no longer makes any sense | |
531 | * with modern CPU's, and this simplifies a lot of things (SMP | |
532 | * and UP become the same). | |
533 | * | |
534 | * NOTE! We used to use the x86 hardware context switching. The | |
535 | * reason for not using it any more becomes apparent when you | |
536 | * try to recover gracefully from saved state that is no longer | |
537 | * valid (stale segment register values in particular). With the | |
538 | * hardware task-switch, there is no way to fix up bad state in | |
539 | * a reasonable manner. | |
540 | * | |
541 | * The fact that Intel documents the hardware task-switching to | |
542 | * be slow is a fairly red herring - this code is not noticeably | |
543 | * faster. However, there _is_ some room for improvement here, | |
544 | * so the performance issues may eventually be a valid point. | |
545 | * More important, however, is the fact that this allows us much | |
546 | * more flexibility. | |
547 | * | |
65ea5b03 | 548 | * The return value (in %ax) will be the "prev" task after |
1da177e4 LT |
549 | * the task-switch, and shows up in ret_from_fork in entry.S, |
550 | * for example. | |
551 | */ | |
8b96f011 FW |
552 | __notrace_funcgraph struct task_struct * |
553 | __switch_to(struct task_struct *prev_p, struct task_struct *next_p) | |
1da177e4 LT |
554 | { |
555 | struct thread_struct *prev = &prev_p->thread, | |
556 | *next = &next_p->thread; | |
557 | int cpu = smp_processor_id(); | |
558 | struct tss_struct *tss = &per_cpu(init_tss, cpu); | |
559 | ||
560 | /* never put a printk in __switch_to... printk() calls wake_up*() indirectly */ | |
561 | ||
562 | __unlazy_fpu(prev_p); | |
563 | ||
acc20761 CE |
564 | |
565 | /* we're going to use this soon, after a few expensive things */ | |
566 | if (next_p->fpu_counter > 5) | |
61c4628b | 567 | prefetch(next->xstate); |
acc20761 | 568 | |
1da177e4 | 569 | /* |
e7a2ff59 | 570 | * Reload esp0. |
1da177e4 | 571 | */ |
faca6227 | 572 | load_sp0(tss, next); |
1da177e4 LT |
573 | |
574 | /* | |
464d1a78 | 575 | * Save away %gs. No need to save %fs, as it was saved on the |
f95d47ca JF |
576 | * stack on entry. No need to save %es and %ds, as those are |
577 | * always kernel segments while inside the kernel. Doing this | |
578 | * before setting the new TLS descriptors avoids the situation | |
579 | * where we temporarily have non-reloadable segments in %fs | |
580 | * and %gs. This could be an issue if the NMI handler ever | |
581 | * used %fs or %gs (it does not today), or if the kernel is | |
582 | * running inside of a hypervisor layer. | |
1da177e4 | 583 | */ |
464d1a78 | 584 | savesegment(gs, prev->gs); |
1da177e4 LT |
585 | |
586 | /* | |
e7a2ff59 | 587 | * Load the per-thread Thread-Local Storage descriptor. |
1da177e4 | 588 | */ |
e7a2ff59 | 589 | load_TLS(next, cpu); |
1da177e4 | 590 | |
8b151144 ZA |
591 | /* |
592 | * Restore IOPL if needed. In normal use, the flags restore | |
593 | * in the switch assembly will handle this. But if the kernel | |
594 | * is running virtualized at a non-zero CPL, the popf will | |
595 | * not restore flags, so it must be done in a separate step. | |
596 | */ | |
597 | if (get_kernel_rpl() && unlikely(prev->iopl != next->iopl)) | |
598 | set_iopl_mask(next->iopl); | |
599 | ||
1da177e4 | 600 | /* |
b3cf2576 | 601 | * Now maybe handle debug registers and/or IO bitmaps |
1da177e4 | 602 | */ |
cf99abac AA |
603 | if (unlikely(task_thread_info(prev_p)->flags & _TIF_WORK_CTXSW_PREV || |
604 | task_thread_info(next_p)->flags & _TIF_WORK_CTXSW_NEXT)) | |
605 | __switch_to_xtra(prev_p, next_p, tss); | |
ffaa8bd6 | 606 | |
9226d125 ZA |
607 | /* |
608 | * Leave lazy mode, flushing any hypercalls made here. | |
609 | * This must be done before restoring TLS segments so | |
610 | * the GDT and LDT are properly updated, and must be | |
611 | * done before math_state_restore, so the TS bit is up | |
612 | * to date. | |
613 | */ | |
614 | arch_leave_lazy_cpu_mode(); | |
615 | ||
acc20761 CE |
616 | /* If the task has used fpu the last 5 timeslices, just do a full |
617 | * restore of the math state immediately to avoid the trap; the | |
618 | * chances of needing FPU soon are obviously high now | |
870568b3 SS |
619 | * |
620 | * tsk_used_math() checks prevent calling math_state_restore(), | |
621 | * which can sleep in the case of !tsk_used_math() | |
acc20761 | 622 | */ |
870568b3 | 623 | if (tsk_used_math(next_p) && next_p->fpu_counter > 5) |
acc20761 CE |
624 | math_state_restore(); |
625 | ||
9226d125 ZA |
626 | /* |
627 | * Restore %gs if needed (which is common) | |
628 | */ | |
629 | if (prev->gs | next->gs) | |
630 | loadsegment(gs, next->gs); | |
631 | ||
7c3576d2 | 632 | x86_write_percpu(current_task, next_p); |
9226d125 | 633 | |
1da177e4 LT |
634 | return prev_p; |
635 | } | |
636 | ||
637 | asmlinkage int sys_fork(struct pt_regs regs) | |
638 | { | |
65ea5b03 | 639 | return do_fork(SIGCHLD, regs.sp, ®s, 0, NULL, NULL); |
1da177e4 LT |
640 | } |
641 | ||
642 | asmlinkage int sys_clone(struct pt_regs regs) | |
643 | { | |
644 | unsigned long clone_flags; | |
645 | unsigned long newsp; | |
646 | int __user *parent_tidptr, *child_tidptr; | |
647 | ||
65ea5b03 PA |
648 | clone_flags = regs.bx; |
649 | newsp = regs.cx; | |
650 | parent_tidptr = (int __user *)regs.dx; | |
651 | child_tidptr = (int __user *)regs.di; | |
1da177e4 | 652 | if (!newsp) |
65ea5b03 | 653 | newsp = regs.sp; |
1da177e4 LT |
654 | return do_fork(clone_flags, newsp, ®s, 0, parent_tidptr, child_tidptr); |
655 | } | |
656 | ||
657 | /* | |
658 | * This is trivial, and on the face of it looks like it | |
659 | * could equally well be done in user mode. | |
660 | * | |
661 | * Not so, for quite unobvious reasons - register pressure. | |
662 | * In user mode vfork() cannot have a stack frame, and if | |
663 | * done by calling the "clone()" system call directly, you | |
664 | * do not have enough call-clobbered registers to hold all | |
665 | * the information you need. | |
666 | */ | |
667 | asmlinkage int sys_vfork(struct pt_regs regs) | |
668 | { | |
65ea5b03 | 669 | return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, regs.sp, ®s, 0, NULL, NULL); |
1da177e4 LT |
670 | } |
671 | ||
672 | /* | |
673 | * sys_execve() executes a new program. | |
674 | */ | |
675 | asmlinkage int sys_execve(struct pt_regs regs) | |
676 | { | |
677 | int error; | |
678 | char * filename; | |
679 | ||
65ea5b03 | 680 | filename = getname((char __user *) regs.bx); |
1da177e4 LT |
681 | error = PTR_ERR(filename); |
682 | if (IS_ERR(filename)) | |
683 | goto out; | |
684 | error = do_execve(filename, | |
65ea5b03 PA |
685 | (char __user * __user *) regs.cx, |
686 | (char __user * __user *) regs.dx, | |
1da177e4 LT |
687 | ®s); |
688 | if (error == 0) { | |
1da177e4 LT |
689 | /* Make sure we don't return using sysenter.. */ |
690 | set_thread_flag(TIF_IRET); | |
691 | } | |
692 | putname(filename); | |
693 | out: | |
694 | return error; | |
695 | } | |
696 | ||
697 | #define top_esp (THREAD_SIZE - sizeof(unsigned long)) | |
698 | #define top_ebp (THREAD_SIZE - 2*sizeof(unsigned long)) | |
699 | ||
700 | unsigned long get_wchan(struct task_struct *p) | |
701 | { | |
65ea5b03 | 702 | unsigned long bp, sp, ip; |
1da177e4 LT |
703 | unsigned long stack_page; |
704 | int count = 0; | |
705 | if (!p || p == current || p->state == TASK_RUNNING) | |
706 | return 0; | |
65e0fdff | 707 | stack_page = (unsigned long)task_stack_page(p); |
faca6227 | 708 | sp = p->thread.sp; |
65ea5b03 | 709 | if (!stack_page || sp < stack_page || sp > top_esp+stack_page) |
1da177e4 | 710 | return 0; |
65ea5b03 PA |
711 | /* include/asm-i386/system.h:switch_to() pushes bp last. */ |
712 | bp = *(unsigned long *) sp; | |
1da177e4 | 713 | do { |
65ea5b03 | 714 | if (bp < stack_page || bp > top_ebp+stack_page) |
1da177e4 | 715 | return 0; |
65ea5b03 PA |
716 | ip = *(unsigned long *) (bp+4); |
717 | if (!in_sched_functions(ip)) | |
718 | return ip; | |
719 | bp = *(unsigned long *) bp; | |
1da177e4 LT |
720 | } while (count++ < 16); |
721 | return 0; | |
722 | } | |
723 | ||
1da177e4 LT |
724 | unsigned long arch_align_stack(unsigned long sp) |
725 | { | |
c16b63e0 | 726 | if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space) |
1da177e4 LT |
727 | sp -= get_random_int() % 8192; |
728 | return sp & ~0xf; | |
729 | } | |
c1d171a0 JK |
730 | |
731 | unsigned long arch_randomize_brk(struct mm_struct *mm) | |
732 | { | |
733 | unsigned long range_end = mm->brk + 0x02000000; | |
734 | return randomize_range(mm->brk, range_end, 0) ? : mm->brk; | |
735 | } |