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> |
befa9e78 JSR |
42 | #include <linux/uaccess.h> |
43 | #include <linux/io.h> | |
44 | #include <linux/kdebug.h> | |
1da177e4 | 45 | |
1da177e4 LT |
46 | #include <asm/pgtable.h> |
47 | #include <asm/system.h> | |
1da177e4 LT |
48 | #include <asm/ldt.h> |
49 | #include <asm/processor.h> | |
50 | #include <asm/i387.h> | |
1da177e4 LT |
51 | #include <asm/desc.h> |
52 | #ifdef CONFIG_MATH_EMULATION | |
53 | #include <asm/math_emu.h> | |
54 | #endif | |
55 | ||
1da177e4 LT |
56 | #include <linux/err.h> |
57 | ||
f3705136 ZM |
58 | #include <asm/tlbflush.h> |
59 | #include <asm/cpu.h> | |
1eda8149 | 60 | #include <asm/idle.h> |
bbc1f698 | 61 | #include <asm/syscalls.h> |
bf53de90 | 62 | #include <asm/ds.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 | ||
1da177e4 LT |
69 | /* |
70 | * Return saved PC of a blocked thread. | |
71 | */ | |
72 | unsigned long thread_saved_pc(struct task_struct *tsk) | |
73 | { | |
faca6227 | 74 | return ((unsigned long *)tsk->thread.sp)[3]; |
1da177e4 LT |
75 | } |
76 | ||
913da64b AN |
77 | #ifndef CONFIG_SMP |
78 | static inline void play_dead(void) | |
79 | { | |
80 | BUG(); | |
81 | } | |
82 | #endif | |
83 | ||
1da177e4 LT |
84 | /* |
85 | * The idle thread. There's no useful work to be | |
86 | * done, so just try to conserve power and have a | |
87 | * low exit latency (ie sit in a loop waiting for | |
88 | * somebody to say that they'd like to reschedule) | |
89 | */ | |
f3705136 | 90 | void cpu_idle(void) |
1da177e4 | 91 | { |
5bfb5d69 | 92 | int cpu = smp_processor_id(); |
f3705136 | 93 | |
495ab9c0 | 94 | current_thread_info()->status |= TS_POLLING; |
64c7c8f8 | 95 | |
1da177e4 LT |
96 | /* endless idle loop with no priority at all */ |
97 | while (1) { | |
b8f8c3cf | 98 | tick_nohz_stop_sched_tick(1); |
1da177e4 | 99 | while (!need_resched()) { |
1da177e4 | 100 | |
f1d1a842 | 101 | check_pgt_cache(); |
1da177e4 | 102 | rmb(); |
1da177e4 | 103 | |
0723a69a BL |
104 | if (rcu_pending(cpu)) |
105 | rcu_check_callbacks(cpu, 0); | |
106 | ||
f3705136 ZM |
107 | if (cpu_is_offline(cpu)) |
108 | play_dead(); | |
109 | ||
7f424a8b | 110 | local_irq_disable(); |
6cd8a4bb SR |
111 | /* Don't trace irqs off for idle */ |
112 | stop_critical_timings(); | |
6ddd2a27 | 113 | pm_idle(); |
6cd8a4bb | 114 | start_critical_timings(); |
1da177e4 | 115 | } |
74167347 | 116 | tick_nohz_restart_sched_tick(); |
5bfb5d69 | 117 | preempt_enable_no_resched(); |
1da177e4 | 118 | schedule(); |
5bfb5d69 | 119 | preempt_disable(); |
1da177e4 LT |
120 | } |
121 | } | |
122 | ||
e2ce07c8 | 123 | void __show_regs(struct pt_regs *regs, int all) |
1da177e4 LT |
124 | { |
125 | unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L; | |
bb1995d5 | 126 | unsigned long d0, d1, d2, d3, d6, d7; |
65ea5b03 | 127 | unsigned long sp; |
9d975ebd | 128 | unsigned short ss, gs; |
90f7d25c | 129 | const char *board; |
9d975ebd PE |
130 | |
131 | if (user_mode_vm(regs)) { | |
65ea5b03 PA |
132 | sp = regs->sp; |
133 | ss = regs->ss & 0xffff; | |
9d975ebd PE |
134 | savesegment(gs, gs); |
135 | } else { | |
65ea5b03 | 136 | sp = (unsigned long) (®s->sp); |
9d975ebd PE |
137 | savesegment(ss, ss); |
138 | savesegment(gs, gs); | |
139 | } | |
1da177e4 LT |
140 | |
141 | printk("\n"); | |
90f7d25c AV |
142 | |
143 | board = dmi_get_system_info(DMI_PRODUCT_NAME); | |
144 | if (!board) | |
145 | board = ""; | |
146 | printk("Pid: %d, comm: %s %s (%s %.*s) %s\n", | |
60812a4a | 147 | task_pid_nr(current), current->comm, |
9d975ebd PE |
148 | print_tainted(), init_utsname()->release, |
149 | (int)strcspn(init_utsname()->version, " "), | |
90f7d25c | 150 | init_utsname()->version, board); |
9d975ebd PE |
151 | |
152 | printk("EIP: %04x:[<%08lx>] EFLAGS: %08lx CPU: %d\n", | |
92bc2056 | 153 | (u16)regs->cs, regs->ip, regs->flags, |
9d975ebd | 154 | smp_processor_id()); |
65ea5b03 | 155 | print_symbol("EIP is at %s\n", regs->ip); |
1da177e4 | 156 | |
1da177e4 | 157 | printk("EAX: %08lx EBX: %08lx ECX: %08lx EDX: %08lx\n", |
65ea5b03 | 158 | regs->ax, regs->bx, regs->cx, regs->dx); |
9d975ebd | 159 | printk("ESI: %08lx EDI: %08lx EBP: %08lx ESP: %08lx\n", |
65ea5b03 | 160 | regs->si, regs->di, regs->bp, sp); |
9d975ebd | 161 | printk(" DS: %04x ES: %04x FS: %04x GS: %04x SS: %04x\n", |
92bc2056 | 162 | (u16)regs->ds, (u16)regs->es, (u16)regs->fs, gs, ss); |
9d975ebd PE |
163 | |
164 | if (!all) | |
165 | return; | |
1da177e4 | 166 | |
4bb0d3ec ZA |
167 | cr0 = read_cr0(); |
168 | cr2 = read_cr2(); | |
169 | cr3 = read_cr3(); | |
ff6e8c0d | 170 | cr4 = read_cr4_safe(); |
9d975ebd PE |
171 | printk("CR0: %08lx CR2: %08lx CR3: %08lx CR4: %08lx\n", |
172 | cr0, cr2, cr3, cr4); | |
bb1995d5 AS |
173 | |
174 | get_debugreg(d0, 0); | |
175 | get_debugreg(d1, 1); | |
176 | get_debugreg(d2, 2); | |
177 | get_debugreg(d3, 3); | |
178 | printk("DR0: %08lx DR1: %08lx DR2: %08lx DR3: %08lx\n", | |
179 | d0, d1, d2, d3); | |
9d975ebd | 180 | |
bb1995d5 AS |
181 | get_debugreg(d6, 6); |
182 | get_debugreg(d7, 7); | |
9d975ebd PE |
183 | printk("DR6: %08lx DR7: %08lx\n", |
184 | d6, d7); | |
185 | } | |
bb1995d5 | 186 | |
9d975ebd PE |
187 | void show_regs(struct pt_regs *regs) |
188 | { | |
e2ce07c8 | 189 | __show_regs(regs, 1); |
5bc27dc2 | 190 | show_trace(NULL, regs, ®s->sp, regs->bp); |
1da177e4 LT |
191 | } |
192 | ||
193 | /* | |
65ea5b03 PA |
194 | * This gets run with %bx containing the |
195 | * function to call, and %dx containing | |
1da177e4 LT |
196 | * the "args". |
197 | */ | |
198 | extern void kernel_thread_helper(void); | |
1da177e4 LT |
199 | |
200 | /* | |
201 | * Create a kernel thread | |
202 | */ | |
befa9e78 | 203 | int kernel_thread(int (*fn)(void *), void *arg, unsigned long flags) |
1da177e4 LT |
204 | { |
205 | struct pt_regs regs; | |
206 | ||
207 | memset(®s, 0, sizeof(regs)); | |
208 | ||
65ea5b03 PA |
209 | regs.bx = (unsigned long) fn; |
210 | regs.dx = (unsigned long) arg; | |
1da177e4 | 211 | |
65ea5b03 PA |
212 | regs.ds = __USER_DS; |
213 | regs.es = __USER_DS; | |
214 | regs.fs = __KERNEL_PERCPU; | |
215 | regs.orig_ax = -1; | |
216 | regs.ip = (unsigned long) kernel_thread_helper; | |
217 | regs.cs = __KERNEL_CS | get_kernel_rpl(); | |
218 | regs.flags = X86_EFLAGS_IF | X86_EFLAGS_SF | X86_EFLAGS_PF | 0x2; | |
1da177e4 LT |
219 | |
220 | /* Ok, create the new process.. */ | |
8cf2c519 | 221 | return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, ®s, 0, NULL, NULL); |
1da177e4 | 222 | } |
129f6946 | 223 | EXPORT_SYMBOL(kernel_thread); |
1da177e4 LT |
224 | |
225 | /* | |
226 | * Free current thread data structures etc.. | |
227 | */ | |
228 | void exit_thread(void) | |
229 | { | |
1da177e4 | 230 | /* The process may have allocated an io port bitmap... nuke it. */ |
b3cf2576 SE |
231 | if (unlikely(test_thread_flag(TIF_IO_BITMAP))) { |
232 | struct task_struct *tsk = current; | |
233 | struct thread_struct *t = &tsk->thread; | |
1da177e4 LT |
234 | int cpu = get_cpu(); |
235 | struct tss_struct *tss = &per_cpu(init_tss, cpu); | |
236 | ||
237 | kfree(t->io_bitmap_ptr); | |
238 | t->io_bitmap_ptr = NULL; | |
b3cf2576 | 239 | clear_thread_flag(TIF_IO_BITMAP); |
1da177e4 LT |
240 | /* |
241 | * Careful, clear this in the TSS too: | |
242 | */ | |
243 | memset(tss->io_bitmap, 0xff, tss->io_bitmap_max); | |
244 | t->io_bitmap_max = 0; | |
245 | tss->io_bitmap_owner = NULL; | |
246 | tss->io_bitmap_max = 0; | |
a75c54f9 | 247 | tss->x86_tss.io_bitmap_base = INVALID_IO_BITMAP_OFFSET; |
1da177e4 LT |
248 | put_cpu(); |
249 | } | |
bf53de90 MM |
250 | |
251 | ds_exit_thread(current); | |
1da177e4 LT |
252 | } |
253 | ||
254 | void flush_thread(void) | |
255 | { | |
256 | struct task_struct *tsk = current; | |
257 | ||
0f534093 RM |
258 | tsk->thread.debugreg0 = 0; |
259 | tsk->thread.debugreg1 = 0; | |
260 | tsk->thread.debugreg2 = 0; | |
261 | tsk->thread.debugreg3 = 0; | |
262 | tsk->thread.debugreg6 = 0; | |
263 | tsk->thread.debugreg7 = 0; | |
befa9e78 | 264 | memset(tsk->thread.tls_array, 0, sizeof(tsk->thread.tls_array)); |
b3cf2576 | 265 | clear_tsk_thread_flag(tsk, TIF_DEBUG); |
1da177e4 LT |
266 | /* |
267 | * Forget coprocessor state.. | |
268 | */ | |
75118a82 | 269 | tsk->fpu_counter = 0; |
1da177e4 LT |
270 | clear_fpu(tsk); |
271 | clear_used_math(); | |
272 | } | |
273 | ||
274 | void release_thread(struct task_struct *dead_task) | |
275 | { | |
2684927c | 276 | BUG_ON(dead_task->mm); |
1da177e4 LT |
277 | release_vm86_irqs(dead_task); |
278 | } | |
279 | ||
280 | /* | |
281 | * This gets called before we allocate a new thread and copy | |
282 | * the current task into it. | |
283 | */ | |
284 | void prepare_to_copy(struct task_struct *tsk) | |
285 | { | |
286 | unlazy_fpu(tsk); | |
287 | } | |
288 | ||
65ea5b03 | 289 | int copy_thread(int nr, unsigned long clone_flags, unsigned long sp, |
1da177e4 | 290 | unsigned long unused, |
befa9e78 | 291 | struct task_struct *p, struct pt_regs *regs) |
1da177e4 | 292 | { |
befa9e78 | 293 | struct pt_regs *childregs; |
1da177e4 LT |
294 | struct task_struct *tsk; |
295 | int err; | |
296 | ||
07b047fc | 297 | childregs = task_pt_regs(p); |
f48d9663 | 298 | *childregs = *regs; |
65ea5b03 PA |
299 | childregs->ax = 0; |
300 | childregs->sp = sp; | |
f48d9663 | 301 | |
faca6227 PA |
302 | p->thread.sp = (unsigned long) childregs; |
303 | p->thread.sp0 = (unsigned long) (childregs+1); | |
1da177e4 | 304 | |
faca6227 | 305 | p->thread.ip = (unsigned long) ret_from_fork; |
1da177e4 | 306 | |
6612538c | 307 | savesegment(gs, p->thread.gs); |
1da177e4 LT |
308 | |
309 | tsk = current; | |
b3cf2576 | 310 | if (unlikely(test_tsk_thread_flag(tsk, TIF_IO_BITMAP))) { |
52978be6 AD |
311 | p->thread.io_bitmap_ptr = kmemdup(tsk->thread.io_bitmap_ptr, |
312 | IO_BITMAP_BYTES, GFP_KERNEL); | |
1da177e4 LT |
313 | if (!p->thread.io_bitmap_ptr) { |
314 | p->thread.io_bitmap_max = 0; | |
315 | return -ENOMEM; | |
316 | } | |
b3cf2576 | 317 | set_tsk_thread_flag(p, TIF_IO_BITMAP); |
1da177e4 LT |
318 | } |
319 | ||
efd1ca52 RM |
320 | err = 0; |
321 | ||
1da177e4 LT |
322 | /* |
323 | * Set a new TLS for the child thread? | |
324 | */ | |
efd1ca52 RM |
325 | if (clone_flags & CLONE_SETTLS) |
326 | err = do_set_thread_area(p, -1, | |
65ea5b03 | 327 | (struct user_desc __user *)childregs->si, 0); |
1da177e4 | 328 | |
1da177e4 LT |
329 | if (err && p->thread.io_bitmap_ptr) { |
330 | kfree(p->thread.io_bitmap_ptr); | |
331 | p->thread.io_bitmap_max = 0; | |
332 | } | |
bf53de90 MM |
333 | |
334 | ds_copy_thread(p, current); | |
335 | ||
336 | clear_tsk_thread_flag(p, TIF_DEBUGCTLMSR); | |
337 | p->thread.debugctlmsr = 0; | |
338 | ||
1da177e4 LT |
339 | return err; |
340 | } | |
341 | ||
513ad84b IM |
342 | void |
343 | start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp) | |
344 | { | |
befa9e78 | 345 | __asm__("movl %0, %%gs" : : "r"(0)); |
513ad84b IM |
346 | regs->fs = 0; |
347 | set_fs(USER_DS); | |
348 | regs->ds = __USER_DS; | |
349 | regs->es = __USER_DS; | |
350 | regs->ss = __USER_DS; | |
351 | regs->cs = __USER_CS; | |
352 | regs->ip = new_ip; | |
353 | regs->sp = new_sp; | |
aa283f49 SS |
354 | /* |
355 | * Free the old FP and other extended state | |
356 | */ | |
357 | free_thread_xstate(current); | |
513ad84b IM |
358 | } |
359 | EXPORT_SYMBOL_GPL(start_thread); | |
360 | ||
bdb4f156 | 361 | static void hard_disable_TSC(void) |
cf99abac AA |
362 | { |
363 | write_cr4(read_cr4() | X86_CR4_TSD); | |
364 | } | |
529e25f6 | 365 | |
cf99abac AA |
366 | void disable_TSC(void) |
367 | { | |
368 | preempt_disable(); | |
369 | if (!test_and_set_thread_flag(TIF_NOTSC)) | |
370 | /* | |
371 | * Must flip the CPU state synchronously with | |
372 | * TIF_NOTSC in the current running context. | |
373 | */ | |
374 | hard_disable_TSC(); | |
375 | preempt_enable(); | |
376 | } | |
529e25f6 | 377 | |
bdb4f156 | 378 | static void hard_enable_TSC(void) |
cf99abac AA |
379 | { |
380 | write_cr4(read_cr4() & ~X86_CR4_TSD); | |
381 | } | |
529e25f6 | 382 | |
a4928cff | 383 | static void enable_TSC(void) |
529e25f6 EB |
384 | { |
385 | preempt_disable(); | |
386 | if (test_and_clear_thread_flag(TIF_NOTSC)) | |
387 | /* | |
388 | * Must flip the CPU state synchronously with | |
389 | * TIF_NOTSC in the current running context. | |
390 | */ | |
391 | hard_enable_TSC(); | |
392 | preempt_enable(); | |
393 | } | |
394 | ||
395 | int get_tsc_mode(unsigned long adr) | |
396 | { | |
397 | unsigned int val; | |
398 | ||
399 | if (test_thread_flag(TIF_NOTSC)) | |
400 | val = PR_TSC_SIGSEGV; | |
401 | else | |
402 | val = PR_TSC_ENABLE; | |
403 | ||
404 | return put_user(val, (unsigned int __user *)adr); | |
405 | } | |
406 | ||
407 | int set_tsc_mode(unsigned int val) | |
408 | { | |
409 | if (val == PR_TSC_SIGSEGV) | |
410 | disable_TSC(); | |
411 | else if (val == PR_TSC_ENABLE) | |
412 | enable_TSC(); | |
413 | else | |
414 | return -EINVAL; | |
415 | ||
416 | return 0; | |
417 | } | |
cf99abac AA |
418 | |
419 | static noinline void | |
420 | __switch_to_xtra(struct task_struct *prev_p, struct task_struct *next_p, | |
421 | struct tss_struct *tss) | |
1da177e4 | 422 | { |
7e991604 | 423 | struct thread_struct *prev, *next; |
b3cf2576 | 424 | |
7e991604 | 425 | prev = &prev_p->thread; |
b3cf2576 SE |
426 | next = &next_p->thread; |
427 | ||
c2724775 MM |
428 | if (test_tsk_thread_flag(next_p, TIF_DS_AREA_MSR) || |
429 | test_tsk_thread_flag(prev_p, TIF_DS_AREA_MSR)) | |
430 | ds_switch_to(prev_p, next_p); | |
431 | else if (next->debugctlmsr != prev->debugctlmsr) | |
5b0e5084 | 432 | update_debugctlmsr(next->debugctlmsr); |
7e991604 | 433 | |
b3cf2576 | 434 | if (test_tsk_thread_flag(next_p, TIF_DEBUG)) { |
0f534093 RM |
435 | set_debugreg(next->debugreg0, 0); |
436 | set_debugreg(next->debugreg1, 1); | |
437 | set_debugreg(next->debugreg2, 2); | |
438 | set_debugreg(next->debugreg3, 3); | |
b3cf2576 | 439 | /* no 4 and 5 */ |
0f534093 RM |
440 | set_debugreg(next->debugreg6, 6); |
441 | set_debugreg(next->debugreg7, 7); | |
b3cf2576 SE |
442 | } |
443 | ||
cf99abac AA |
444 | if (test_tsk_thread_flag(prev_p, TIF_NOTSC) ^ |
445 | test_tsk_thread_flag(next_p, TIF_NOTSC)) { | |
446 | /* prev and next are different */ | |
447 | if (test_tsk_thread_flag(next_p, TIF_NOTSC)) | |
448 | hard_disable_TSC(); | |
449 | else | |
450 | hard_enable_TSC(); | |
451 | } | |
cf99abac | 452 | |
b3cf2576 | 453 | if (!test_tsk_thread_flag(next_p, TIF_IO_BITMAP)) { |
1da177e4 LT |
454 | /* |
455 | * Disable the bitmap via an invalid offset. We still cache | |
456 | * the previous bitmap owner and the IO bitmap contents: | |
457 | */ | |
a75c54f9 | 458 | tss->x86_tss.io_bitmap_base = INVALID_IO_BITMAP_OFFSET; |
1da177e4 LT |
459 | return; |
460 | } | |
b3cf2576 | 461 | |
1da177e4 LT |
462 | if (likely(next == tss->io_bitmap_owner)) { |
463 | /* | |
464 | * Previous owner of the bitmap (hence the bitmap content) | |
465 | * matches the next task, we dont have to do anything but | |
466 | * to set a valid offset in the TSS: | |
467 | */ | |
a75c54f9 | 468 | tss->x86_tss.io_bitmap_base = IO_BITMAP_OFFSET; |
1da177e4 LT |
469 | return; |
470 | } | |
471 | /* | |
472 | * Lazy TSS's I/O bitmap copy. We set an invalid offset here | |
473 | * and we let the task to get a GPF in case an I/O instruction | |
474 | * is performed. The handler of the GPF will verify that the | |
475 | * faulting task has a valid I/O bitmap and, it true, does the | |
476 | * real copy and restart the instruction. This will save us | |
477 | * redundant copies when the currently switched task does not | |
478 | * perform any I/O during its timeslice. | |
479 | */ | |
a75c54f9 | 480 | tss->x86_tss.io_bitmap_base = INVALID_IO_BITMAP_OFFSET_LAZY; |
1da177e4 | 481 | } |
1da177e4 LT |
482 | |
483 | /* | |
484 | * switch_to(x,yn) should switch tasks from x to y. | |
485 | * | |
486 | * We fsave/fwait so that an exception goes off at the right time | |
487 | * (as a call from the fsave or fwait in effect) rather than to | |
488 | * the wrong process. Lazy FP saving no longer makes any sense | |
489 | * with modern CPU's, and this simplifies a lot of things (SMP | |
490 | * and UP become the same). | |
491 | * | |
492 | * NOTE! We used to use the x86 hardware context switching. The | |
493 | * reason for not using it any more becomes apparent when you | |
494 | * try to recover gracefully from saved state that is no longer | |
495 | * valid (stale segment register values in particular). With the | |
496 | * hardware task-switch, there is no way to fix up bad state in | |
497 | * a reasonable manner. | |
498 | * | |
499 | * The fact that Intel documents the hardware task-switching to | |
500 | * be slow is a fairly red herring - this code is not noticeably | |
501 | * faster. However, there _is_ some room for improvement here, | |
502 | * so the performance issues may eventually be a valid point. | |
503 | * More important, however, is the fact that this allows us much | |
504 | * more flexibility. | |
505 | * | |
65ea5b03 | 506 | * The return value (in %ax) will be the "prev" task after |
1da177e4 LT |
507 | * the task-switch, and shows up in ret_from_fork in entry.S, |
508 | * for example. | |
509 | */ | |
8b96f011 FW |
510 | __notrace_funcgraph struct task_struct * |
511 | __switch_to(struct task_struct *prev_p, struct task_struct *next_p) | |
1da177e4 LT |
512 | { |
513 | struct thread_struct *prev = &prev_p->thread, | |
514 | *next = &next_p->thread; | |
515 | int cpu = smp_processor_id(); | |
516 | struct tss_struct *tss = &per_cpu(init_tss, cpu); | |
517 | ||
518 | /* never put a printk in __switch_to... printk() calls wake_up*() indirectly */ | |
519 | ||
520 | __unlazy_fpu(prev_p); | |
521 | ||
acc20761 CE |
522 | |
523 | /* we're going to use this soon, after a few expensive things */ | |
524 | if (next_p->fpu_counter > 5) | |
61c4628b | 525 | prefetch(next->xstate); |
acc20761 | 526 | |
1da177e4 | 527 | /* |
e7a2ff59 | 528 | * Reload esp0. |
1da177e4 | 529 | */ |
faca6227 | 530 | load_sp0(tss, next); |
1da177e4 LT |
531 | |
532 | /* | |
464d1a78 | 533 | * Save away %gs. No need to save %fs, as it was saved on the |
f95d47ca JF |
534 | * stack on entry. No need to save %es and %ds, as those are |
535 | * always kernel segments while inside the kernel. Doing this | |
536 | * before setting the new TLS descriptors avoids the situation | |
537 | * where we temporarily have non-reloadable segments in %fs | |
538 | * and %gs. This could be an issue if the NMI handler ever | |
539 | * used %fs or %gs (it does not today), or if the kernel is | |
540 | * running inside of a hypervisor layer. | |
1da177e4 | 541 | */ |
464d1a78 | 542 | savesegment(gs, prev->gs); |
1da177e4 LT |
543 | |
544 | /* | |
e7a2ff59 | 545 | * Load the per-thread Thread-Local Storage descriptor. |
1da177e4 | 546 | */ |
e7a2ff59 | 547 | load_TLS(next, cpu); |
1da177e4 | 548 | |
8b151144 ZA |
549 | /* |
550 | * Restore IOPL if needed. In normal use, the flags restore | |
551 | * in the switch assembly will handle this. But if the kernel | |
552 | * is running virtualized at a non-zero CPL, the popf will | |
553 | * not restore flags, so it must be done in a separate step. | |
554 | */ | |
555 | if (get_kernel_rpl() && unlikely(prev->iopl != next->iopl)) | |
556 | set_iopl_mask(next->iopl); | |
557 | ||
1da177e4 | 558 | /* |
b3cf2576 | 559 | * Now maybe handle debug registers and/or IO bitmaps |
1da177e4 | 560 | */ |
cf99abac AA |
561 | if (unlikely(task_thread_info(prev_p)->flags & _TIF_WORK_CTXSW_PREV || |
562 | task_thread_info(next_p)->flags & _TIF_WORK_CTXSW_NEXT)) | |
563 | __switch_to_xtra(prev_p, next_p, tss); | |
ffaa8bd6 | 564 | |
9226d125 ZA |
565 | /* |
566 | * Leave lazy mode, flushing any hypercalls made here. | |
567 | * This must be done before restoring TLS segments so | |
568 | * the GDT and LDT are properly updated, and must be | |
569 | * done before math_state_restore, so the TS bit is up | |
570 | * to date. | |
571 | */ | |
572 | arch_leave_lazy_cpu_mode(); | |
573 | ||
acc20761 CE |
574 | /* If the task has used fpu the last 5 timeslices, just do a full |
575 | * restore of the math state immediately to avoid the trap; the | |
576 | * chances of needing FPU soon are obviously high now | |
870568b3 SS |
577 | * |
578 | * tsk_used_math() checks prevent calling math_state_restore(), | |
579 | * which can sleep in the case of !tsk_used_math() | |
acc20761 | 580 | */ |
870568b3 | 581 | if (tsk_used_math(next_p) && next_p->fpu_counter > 5) |
acc20761 CE |
582 | math_state_restore(); |
583 | ||
9226d125 ZA |
584 | /* |
585 | * Restore %gs if needed (which is common) | |
586 | */ | |
587 | if (prev->gs | next->gs) | |
588 | loadsegment(gs, next->gs); | |
589 | ||
6dbde353 | 590 | percpu_write(current_task, next_p); |
9226d125 | 591 | |
1da177e4 LT |
592 | return prev_p; |
593 | } | |
594 | ||
595 | asmlinkage int sys_fork(struct pt_regs regs) | |
596 | { | |
65ea5b03 | 597 | return do_fork(SIGCHLD, regs.sp, ®s, 0, NULL, NULL); |
1da177e4 LT |
598 | } |
599 | ||
600 | asmlinkage int sys_clone(struct pt_regs regs) | |
601 | { | |
602 | unsigned long clone_flags; | |
603 | unsigned long newsp; | |
604 | int __user *parent_tidptr, *child_tidptr; | |
605 | ||
65ea5b03 PA |
606 | clone_flags = regs.bx; |
607 | newsp = regs.cx; | |
608 | parent_tidptr = (int __user *)regs.dx; | |
609 | child_tidptr = (int __user *)regs.di; | |
1da177e4 | 610 | if (!newsp) |
65ea5b03 | 611 | newsp = regs.sp; |
1da177e4 LT |
612 | return do_fork(clone_flags, newsp, ®s, 0, parent_tidptr, child_tidptr); |
613 | } | |
614 | ||
615 | /* | |
616 | * This is trivial, and on the face of it looks like it | |
617 | * could equally well be done in user mode. | |
618 | * | |
619 | * Not so, for quite unobvious reasons - register pressure. | |
620 | * In user mode vfork() cannot have a stack frame, and if | |
621 | * done by calling the "clone()" system call directly, you | |
622 | * do not have enough call-clobbered registers to hold all | |
623 | * the information you need. | |
624 | */ | |
625 | asmlinkage int sys_vfork(struct pt_regs regs) | |
626 | { | |
65ea5b03 | 627 | return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, regs.sp, ®s, 0, NULL, NULL); |
1da177e4 LT |
628 | } |
629 | ||
630 | /* | |
631 | * sys_execve() executes a new program. | |
632 | */ | |
633 | asmlinkage int sys_execve(struct pt_regs regs) | |
634 | { | |
635 | int error; | |
befa9e78 | 636 | char *filename; |
1da177e4 | 637 | |
65ea5b03 | 638 | filename = getname((char __user *) regs.bx); |
1da177e4 LT |
639 | error = PTR_ERR(filename); |
640 | if (IS_ERR(filename)) | |
641 | goto out; | |
642 | error = do_execve(filename, | |
65ea5b03 PA |
643 | (char __user * __user *) regs.cx, |
644 | (char __user * __user *) regs.dx, | |
1da177e4 LT |
645 | ®s); |
646 | if (error == 0) { | |
1da177e4 LT |
647 | /* Make sure we don't return using sysenter.. */ |
648 | set_thread_flag(TIF_IRET); | |
649 | } | |
650 | putname(filename); | |
651 | out: | |
652 | return error; | |
653 | } | |
654 | ||
655 | #define top_esp (THREAD_SIZE - sizeof(unsigned long)) | |
656 | #define top_ebp (THREAD_SIZE - 2*sizeof(unsigned long)) | |
657 | ||
658 | unsigned long get_wchan(struct task_struct *p) | |
659 | { | |
65ea5b03 | 660 | unsigned long bp, sp, ip; |
1da177e4 LT |
661 | unsigned long stack_page; |
662 | int count = 0; | |
663 | if (!p || p == current || p->state == TASK_RUNNING) | |
664 | return 0; | |
65e0fdff | 665 | stack_page = (unsigned long)task_stack_page(p); |
faca6227 | 666 | sp = p->thread.sp; |
65ea5b03 | 667 | if (!stack_page || sp < stack_page || sp > top_esp+stack_page) |
1da177e4 | 668 | return 0; |
65ea5b03 PA |
669 | /* include/asm-i386/system.h:switch_to() pushes bp last. */ |
670 | bp = *(unsigned long *) sp; | |
1da177e4 | 671 | do { |
65ea5b03 | 672 | if (bp < stack_page || bp > top_ebp+stack_page) |
1da177e4 | 673 | return 0; |
65ea5b03 PA |
674 | ip = *(unsigned long *) (bp+4); |
675 | if (!in_sched_functions(ip)) | |
676 | return ip; | |
677 | bp = *(unsigned long *) bp; | |
1da177e4 LT |
678 | } while (count++ < 16); |
679 | return 0; | |
680 | } | |
681 | ||
1da177e4 LT |
682 | unsigned long arch_align_stack(unsigned long sp) |
683 | { | |
c16b63e0 | 684 | if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space) |
1da177e4 LT |
685 | sp -= get_random_int() % 8192; |
686 | return sp & ~0xf; | |
687 | } | |
c1d171a0 JK |
688 | |
689 | unsigned long arch_randomize_brk(struct mm_struct *mm) | |
690 | { | |
691 | unsigned long range_end = mm->brk + 0x02000000; | |
692 | return randomize_range(mm->brk, range_end, 0) ? : mm->brk; | |
693 | } |