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 | |
6612538c | 6 | * |
1da177e4 LT |
7 | * X86-64 port |
8 | * Andi Kleen. | |
76e4f660 AR |
9 | * |
10 | * CPU hotplug support - ashok.raj@intel.com | |
1da177e4 LT |
11 | */ |
12 | ||
13 | /* | |
14 | * This file handles the architecture-dependent parts of process handling.. | |
15 | */ | |
16 | ||
17 | #include <stdarg.h> | |
18 | ||
76e4f660 | 19 | #include <linux/cpu.h> |
1da177e4 LT |
20 | #include <linux/errno.h> |
21 | #include <linux/sched.h> | |
6612538c | 22 | #include <linux/fs.h> |
1da177e4 LT |
23 | #include <linux/kernel.h> |
24 | #include <linux/mm.h> | |
25 | #include <linux/elfcore.h> | |
26 | #include <linux/smp.h> | |
27 | #include <linux/slab.h> | |
28 | #include <linux/user.h> | |
1da177e4 | 29 | #include <linux/interrupt.h> |
6612538c | 30 | #include <linux/utsname.h> |
1da177e4 | 31 | #include <linux/delay.h> |
6612538c | 32 | #include <linux/module.h> |
1da177e4 | 33 | #include <linux/ptrace.h> |
1da177e4 | 34 | #include <linux/random.h> |
95833c83 | 35 | #include <linux/notifier.h> |
c6fd91f0 | 36 | #include <linux/kprobes.h> |
1eeb66a1 | 37 | #include <linux/kdebug.h> |
02290683 | 38 | #include <linux/tick.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/processor.h> | |
46 | #include <asm/i387.h> | |
47 | #include <asm/mmu_context.h> | |
48 | #include <asm/pda.h> | |
49 | #include <asm/prctl.h> | |
1da177e4 LT |
50 | #include <asm/desc.h> |
51 | #include <asm/proto.h> | |
52 | #include <asm/ia32.h> | |
95833c83 | 53 | #include <asm/idle.h> |
1da177e4 LT |
54 | |
55 | asmlinkage extern void ret_from_fork(void); | |
56 | ||
57 | unsigned long kernel_thread_flags = CLONE_VM | CLONE_UNTRACED; | |
58 | ||
e041c683 | 59 | static ATOMIC_NOTIFIER_HEAD(idle_notifier); |
95833c83 AK |
60 | |
61 | void idle_notifier_register(struct notifier_block *n) | |
62 | { | |
e041c683 | 63 | atomic_notifier_chain_register(&idle_notifier, n); |
95833c83 | 64 | } |
95833c83 | 65 | |
95833c83 AK |
66 | void enter_idle(void) |
67 | { | |
a15da49d | 68 | write_pda(isidle, 1); |
e041c683 | 69 | atomic_notifier_call_chain(&idle_notifier, IDLE_START, NULL); |
95833c83 AK |
70 | } |
71 | ||
72 | static void __exit_idle(void) | |
73 | { | |
9446868b | 74 | if (test_and_clear_bit_pda(0, isidle) == 0) |
a15da49d | 75 | return; |
e041c683 | 76 | atomic_notifier_call_chain(&idle_notifier, IDLE_END, NULL); |
95833c83 AK |
77 | } |
78 | ||
79 | /* Called from interrupts to signify idle end */ | |
80 | void exit_idle(void) | |
81 | { | |
a15da49d AK |
82 | /* idle loop has pid 0 */ |
83 | if (current->pid) | |
95833c83 AK |
84 | return; |
85 | __exit_idle(); | |
86 | } | |
87 | ||
76e4f660 AR |
88 | #ifdef CONFIG_HOTPLUG_CPU |
89 | DECLARE_PER_CPU(int, cpu_state); | |
90 | ||
91 | #include <asm/nmi.h> | |
1fa744e6 | 92 | /* We halt the CPU with physical CPU hotplug */ |
76e4f660 AR |
93 | static inline void play_dead(void) |
94 | { | |
95 | idle_task_exit(); | |
96 | wbinvd(); | |
97 | mb(); | |
98 | /* Ack it */ | |
99 | __get_cpu_var(cpu_state) = CPU_DEAD; | |
100 | ||
1fa744e6 | 101 | local_irq_disable(); |
76e4f660 | 102 | while (1) |
1fa744e6 | 103 | halt(); |
76e4f660 AR |
104 | } |
105 | #else | |
106 | static inline void play_dead(void) | |
107 | { | |
108 | BUG(); | |
109 | } | |
110 | #endif /* CONFIG_HOTPLUG_CPU */ | |
111 | ||
1da177e4 LT |
112 | /* |
113 | * The idle thread. There's no useful work to be | |
114 | * done, so just try to conserve power and have a | |
115 | * low exit latency (ie sit in a loop waiting for | |
116 | * somebody to say that they'd like to reschedule) | |
117 | */ | |
b10db7f0 | 118 | void cpu_idle(void) |
1da177e4 | 119 | { |
495ab9c0 | 120 | current_thread_info()->status |= TS_POLLING; |
1da177e4 LT |
121 | /* endless idle loop with no priority at all */ |
122 | while (1) { | |
b8f8c3cf | 123 | tick_nohz_stop_sched_tick(1); |
1da177e4 | 124 | while (!need_resched()) { |
1da177e4 | 125 | |
1da177e4 | 126 | rmb(); |
6ddd2a27 | 127 | |
76e4f660 AR |
128 | if (cpu_is_offline(smp_processor_id())) |
129 | play_dead(); | |
d331e739 VP |
130 | /* |
131 | * Idle routines should keep interrupts disabled | |
132 | * from here on, until they go to idle. | |
133 | * Otherwise, idle callbacks can misfire. | |
134 | */ | |
135 | local_irq_disable(); | |
95833c83 | 136 | enter_idle(); |
81d68a96 SR |
137 | /* Don't trace irqs off for idle */ |
138 | stop_critical_timings(); | |
6ddd2a27 | 139 | pm_idle(); |
81d68a96 | 140 | start_critical_timings(); |
a15da49d AK |
141 | /* In many cases the interrupt that ended idle |
142 | has already called exit_idle. But some idle | |
143 | loops can be woken up without interrupt. */ | |
95833c83 | 144 | __exit_idle(); |
1da177e4 LT |
145 | } |
146 | ||
02290683 | 147 | tick_nohz_restart_sched_tick(); |
5bfb5d69 | 148 | preempt_enable_no_resched(); |
1da177e4 | 149 | schedule(); |
5bfb5d69 | 150 | preempt_disable(); |
1da177e4 LT |
151 | } |
152 | } | |
153 | ||
6612538c | 154 | /* Prints also some state that isn't saved in the pt_regs */ |
1da177e4 LT |
155 | void __show_regs(struct pt_regs * regs) |
156 | { | |
157 | unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L, fs, gs, shadowgs; | |
bb1995d5 | 158 | unsigned long d0, d1, d2, d3, d6, d7; |
6612538c HS |
159 | unsigned int fsindex, gsindex; |
160 | unsigned int ds, cs, es; | |
1da177e4 LT |
161 | |
162 | printk("\n"); | |
163 | print_modules(); | |
9acf23c4 AK |
164 | printk("Pid: %d, comm: %.20s %s %s %.*s\n", |
165 | current->pid, current->comm, print_tainted(), | |
96b644bd SH |
166 | init_utsname()->release, |
167 | (int)strcspn(init_utsname()->version, " "), | |
168 | init_utsname()->version); | |
65ea5b03 | 169 | printk("RIP: %04lx:[<%016lx>] ", regs->cs & 0xffff, regs->ip); |
aafbd7eb | 170 | printk_address(regs->ip, 1); |
65ea5b03 PA |
171 | printk("RSP: %04lx:%016lx EFLAGS: %08lx\n", regs->ss, regs->sp, |
172 | regs->flags); | |
1da177e4 | 173 | printk("RAX: %016lx RBX: %016lx RCX: %016lx\n", |
65ea5b03 | 174 | regs->ax, regs->bx, regs->cx); |
1da177e4 | 175 | printk("RDX: %016lx RSI: %016lx RDI: %016lx\n", |
65ea5b03 | 176 | regs->dx, regs->si, regs->di); |
1da177e4 | 177 | printk("RBP: %016lx R08: %016lx R09: %016lx\n", |
65ea5b03 | 178 | regs->bp, regs->r8, regs->r9); |
1da177e4 LT |
179 | printk("R10: %016lx R11: %016lx R12: %016lx\n", |
180 | regs->r10, regs->r11, regs->r12); | |
181 | printk("R13: %016lx R14: %016lx R15: %016lx\n", | |
182 | regs->r13, regs->r14, regs->r15); | |
183 | ||
184 | asm("movl %%ds,%0" : "=r" (ds)); | |
185 | asm("movl %%cs,%0" : "=r" (cs)); | |
186 | asm("movl %%es,%0" : "=r" (es)); | |
187 | asm("movl %%fs,%0" : "=r" (fsindex)); | |
188 | asm("movl %%gs,%0" : "=r" (gsindex)); | |
189 | ||
190 | rdmsrl(MSR_FS_BASE, fs); | |
191 | rdmsrl(MSR_GS_BASE, gs); | |
192 | rdmsrl(MSR_KERNEL_GS_BASE, shadowgs); | |
193 | ||
f51c9452 GOC |
194 | cr0 = read_cr0(); |
195 | cr2 = read_cr2(); | |
196 | cr3 = read_cr3(); | |
197 | cr4 = read_cr4(); | |
1da177e4 LT |
198 | |
199 | printk("FS: %016lx(%04x) GS:%016lx(%04x) knlGS:%016lx\n", | |
200 | fs,fsindex,gs,gsindex,shadowgs); | |
201 | printk("CS: %04x DS: %04x ES: %04x CR0: %016lx\n", cs, ds, es, cr0); | |
202 | printk("CR2: %016lx CR3: %016lx CR4: %016lx\n", cr2, cr3, cr4); | |
bb1995d5 AS |
203 | |
204 | get_debugreg(d0, 0); | |
205 | get_debugreg(d1, 1); | |
206 | get_debugreg(d2, 2); | |
207 | printk("DR0: %016lx DR1: %016lx DR2: %016lx\n", d0, d1, d2); | |
208 | get_debugreg(d3, 3); | |
209 | get_debugreg(d6, 6); | |
210 | get_debugreg(d7, 7); | |
211 | printk("DR3: %016lx DR6: %016lx DR7: %016lx\n", d3, d6, d7); | |
1da177e4 LT |
212 | } |
213 | ||
214 | void show_regs(struct pt_regs *regs) | |
215 | { | |
c078d326 | 216 | printk("CPU %d:", smp_processor_id()); |
1da177e4 | 217 | __show_regs(regs); |
bc850d6b | 218 | show_trace(NULL, regs, (void *)(regs + 1), regs->bp); |
1da177e4 LT |
219 | } |
220 | ||
221 | /* | |
222 | * Free current thread data structures etc.. | |
223 | */ | |
224 | void exit_thread(void) | |
225 | { | |
226 | struct task_struct *me = current; | |
227 | struct thread_struct *t = &me->thread; | |
73649dab | 228 | |
6612538c | 229 | if (me->thread.io_bitmap_ptr) { |
1da177e4 LT |
230 | struct tss_struct *tss = &per_cpu(init_tss, get_cpu()); |
231 | ||
232 | kfree(t->io_bitmap_ptr); | |
233 | t->io_bitmap_ptr = NULL; | |
d3a4f48d | 234 | clear_thread_flag(TIF_IO_BITMAP); |
1da177e4 LT |
235 | /* |
236 | * Careful, clear this in the TSS too: | |
237 | */ | |
238 | memset(tss->io_bitmap, 0xff, t->io_bitmap_max); | |
239 | t->io_bitmap_max = 0; | |
240 | put_cpu(); | |
241 | } | |
242 | } | |
243 | ||
244 | void flush_thread(void) | |
245 | { | |
246 | struct task_struct *tsk = current; | |
1da177e4 | 247 | |
303cd153 MD |
248 | if (test_tsk_thread_flag(tsk, TIF_ABI_PENDING)) { |
249 | clear_tsk_thread_flag(tsk, TIF_ABI_PENDING); | |
250 | if (test_tsk_thread_flag(tsk, TIF_IA32)) { | |
251 | clear_tsk_thread_flag(tsk, TIF_IA32); | |
252 | } else { | |
253 | set_tsk_thread_flag(tsk, TIF_IA32); | |
4d9bc79c | 254 | current_thread_info()->status |= TS_COMPAT; |
303cd153 | 255 | } |
4d9bc79c | 256 | } |
303cd153 | 257 | clear_tsk_thread_flag(tsk, TIF_DEBUG); |
1da177e4 LT |
258 | |
259 | tsk->thread.debugreg0 = 0; | |
260 | tsk->thread.debugreg1 = 0; | |
261 | tsk->thread.debugreg2 = 0; | |
262 | tsk->thread.debugreg3 = 0; | |
263 | tsk->thread.debugreg6 = 0; | |
264 | tsk->thread.debugreg7 = 0; | |
6612538c | 265 | memset(tsk->thread.tls_array, 0, sizeof(tsk->thread.tls_array)); |
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 | { | |
276 | if (dead_task->mm) { | |
277 | if (dead_task->mm->context.size) { | |
278 | printk("WARNING: dead process %8s still has LDT? <%p/%d>\n", | |
279 | dead_task->comm, | |
280 | dead_task->mm->context.ldt, | |
281 | dead_task->mm->context.size); | |
282 | BUG(); | |
283 | } | |
284 | } | |
285 | } | |
286 | ||
287 | static inline void set_32bit_tls(struct task_struct *t, int tls, u32 addr) | |
288 | { | |
6612538c | 289 | struct user_desc ud = { |
1da177e4 LT |
290 | .base_addr = addr, |
291 | .limit = 0xfffff, | |
292 | .seg_32bit = 1, | |
293 | .limit_in_pages = 1, | |
294 | .useable = 1, | |
295 | }; | |
ade1af77 | 296 | struct desc_struct *desc = t->thread.tls_array; |
1da177e4 | 297 | desc += tls; |
80fbb69a | 298 | fill_ldt(desc, &ud); |
1da177e4 LT |
299 | } |
300 | ||
301 | static inline u32 read_32bit_tls(struct task_struct *t, int tls) | |
302 | { | |
91394eb0 | 303 | return get_desc_base(&t->thread.tls_array[tls]); |
1da177e4 LT |
304 | } |
305 | ||
306 | /* | |
307 | * This gets called before we allocate a new thread and copy | |
308 | * the current task into it. | |
309 | */ | |
310 | void prepare_to_copy(struct task_struct *tsk) | |
311 | { | |
312 | unlazy_fpu(tsk); | |
313 | } | |
314 | ||
65ea5b03 | 315 | int copy_thread(int nr, unsigned long clone_flags, unsigned long sp, |
1da177e4 LT |
316 | unsigned long unused, |
317 | struct task_struct * p, struct pt_regs * regs) | |
318 | { | |
319 | int err; | |
320 | struct pt_regs * childregs; | |
321 | struct task_struct *me = current; | |
322 | ||
a88cde13 | 323 | childregs = ((struct pt_regs *) |
57eafdc2 | 324 | (THREAD_SIZE + task_stack_page(p))) - 1; |
1da177e4 LT |
325 | *childregs = *regs; |
326 | ||
65ea5b03 PA |
327 | childregs->ax = 0; |
328 | childregs->sp = sp; | |
329 | if (sp == ~0UL) | |
330 | childregs->sp = (unsigned long)childregs; | |
1da177e4 | 331 | |
faca6227 PA |
332 | p->thread.sp = (unsigned long) childregs; |
333 | p->thread.sp0 = (unsigned long) (childregs+1); | |
334 | p->thread.usersp = me->thread.usersp; | |
1da177e4 | 335 | |
e4f17c43 | 336 | set_tsk_thread_flag(p, TIF_FORK); |
1da177e4 LT |
337 | |
338 | p->thread.fs = me->thread.fs; | |
339 | p->thread.gs = me->thread.gs; | |
340 | ||
ada85708 JF |
341 | savesegment(gs, p->thread.gsindex); |
342 | savesegment(fs, p->thread.fsindex); | |
343 | savesegment(es, p->thread.es); | |
344 | savesegment(ds, p->thread.ds); | |
1da177e4 | 345 | |
d3a4f48d | 346 | if (unlikely(test_tsk_thread_flag(me, TIF_IO_BITMAP))) { |
1da177e4 LT |
347 | p->thread.io_bitmap_ptr = kmalloc(IO_BITMAP_BYTES, GFP_KERNEL); |
348 | if (!p->thread.io_bitmap_ptr) { | |
349 | p->thread.io_bitmap_max = 0; | |
350 | return -ENOMEM; | |
351 | } | |
a88cde13 AK |
352 | memcpy(p->thread.io_bitmap_ptr, me->thread.io_bitmap_ptr, |
353 | IO_BITMAP_BYTES); | |
d3a4f48d | 354 | set_tsk_thread_flag(p, TIF_IO_BITMAP); |
6612538c | 355 | } |
1da177e4 LT |
356 | |
357 | /* | |
358 | * Set a new TLS for the child thread? | |
359 | */ | |
360 | if (clone_flags & CLONE_SETTLS) { | |
361 | #ifdef CONFIG_IA32_EMULATION | |
362 | if (test_thread_flag(TIF_IA32)) | |
efd1ca52 | 363 | err = do_set_thread_area(p, -1, |
65ea5b03 | 364 | (struct user_desc __user *)childregs->si, 0); |
1da177e4 LT |
365 | else |
366 | #endif | |
367 | err = do_arch_prctl(p, ARCH_SET_FS, childregs->r8); | |
368 | if (err) | |
369 | goto out; | |
370 | } | |
371 | err = 0; | |
372 | out: | |
373 | if (err && p->thread.io_bitmap_ptr) { | |
374 | kfree(p->thread.io_bitmap_ptr); | |
375 | p->thread.io_bitmap_max = 0; | |
376 | } | |
377 | return err; | |
378 | } | |
379 | ||
513ad84b IM |
380 | void |
381 | start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp) | |
382 | { | |
ada85708 JF |
383 | loadsegment(fs, 0); |
384 | loadsegment(es, 0); | |
385 | loadsegment(ds, 0); | |
513ad84b IM |
386 | load_gs_index(0); |
387 | regs->ip = new_ip; | |
388 | regs->sp = new_sp; | |
389 | write_pda(oldrsp, new_sp); | |
390 | regs->cs = __USER_CS; | |
391 | regs->ss = __USER_DS; | |
392 | regs->flags = 0x200; | |
393 | set_fs(USER_DS); | |
aa283f49 SS |
394 | /* |
395 | * Free the old FP and other extended state | |
396 | */ | |
397 | free_thread_xstate(current); | |
513ad84b IM |
398 | } |
399 | EXPORT_SYMBOL_GPL(start_thread); | |
400 | ||
529e25f6 EB |
401 | static void hard_disable_TSC(void) |
402 | { | |
403 | write_cr4(read_cr4() | X86_CR4_TSD); | |
404 | } | |
405 | ||
406 | void disable_TSC(void) | |
407 | { | |
408 | preempt_disable(); | |
409 | if (!test_and_set_thread_flag(TIF_NOTSC)) | |
410 | /* | |
411 | * Must flip the CPU state synchronously with | |
412 | * TIF_NOTSC in the current running context. | |
413 | */ | |
414 | hard_disable_TSC(); | |
415 | preempt_enable(); | |
416 | } | |
417 | ||
418 | static void hard_enable_TSC(void) | |
419 | { | |
420 | write_cr4(read_cr4() & ~X86_CR4_TSD); | |
421 | } | |
422 | ||
a4928cff | 423 | static void enable_TSC(void) |
529e25f6 EB |
424 | { |
425 | preempt_disable(); | |
426 | if (test_and_clear_thread_flag(TIF_NOTSC)) | |
427 | /* | |
428 | * Must flip the CPU state synchronously with | |
429 | * TIF_NOTSC in the current running context. | |
430 | */ | |
431 | hard_enable_TSC(); | |
432 | preempt_enable(); | |
433 | } | |
434 | ||
435 | int get_tsc_mode(unsigned long adr) | |
436 | { | |
437 | unsigned int val; | |
438 | ||
439 | if (test_thread_flag(TIF_NOTSC)) | |
440 | val = PR_TSC_SIGSEGV; | |
441 | else | |
442 | val = PR_TSC_ENABLE; | |
443 | ||
444 | return put_user(val, (unsigned int __user *)adr); | |
445 | } | |
446 | ||
447 | int set_tsc_mode(unsigned int val) | |
448 | { | |
449 | if (val == PR_TSC_SIGSEGV) | |
450 | disable_TSC(); | |
451 | else if (val == PR_TSC_ENABLE) | |
452 | enable_TSC(); | |
453 | else | |
454 | return -EINVAL; | |
455 | ||
456 | return 0; | |
457 | } | |
458 | ||
1da177e4 LT |
459 | /* |
460 | * This special macro can be used to load a debugging register | |
461 | */ | |
6612538c HS |
462 | #define loaddebug(thread, r) set_debugreg(thread->debugreg ## r, r) |
463 | ||
d3a4f48d | 464 | static inline void __switch_to_xtra(struct task_struct *prev_p, |
6612538c HS |
465 | struct task_struct *next_p, |
466 | struct tss_struct *tss) | |
d3a4f48d SE |
467 | { |
468 | struct thread_struct *prev, *next; | |
eee3af4a | 469 | unsigned long debugctl; |
d3a4f48d SE |
470 | |
471 | prev = &prev_p->thread, | |
472 | next = &next_p->thread; | |
473 | ||
eee3af4a MM |
474 | debugctl = prev->debugctlmsr; |
475 | if (next->ds_area_msr != prev->ds_area_msr) { | |
476 | /* we clear debugctl to make sure DS | |
477 | * is not in use when we change it */ | |
478 | debugctl = 0; | |
5b0e5084 | 479 | update_debugctlmsr(0); |
eee3af4a MM |
480 | wrmsrl(MSR_IA32_DS_AREA, next->ds_area_msr); |
481 | } | |
482 | ||
483 | if (next->debugctlmsr != debugctl) | |
5b0e5084 | 484 | update_debugctlmsr(next->debugctlmsr); |
7e991604 | 485 | |
d3a4f48d SE |
486 | if (test_tsk_thread_flag(next_p, TIF_DEBUG)) { |
487 | loaddebug(next, 0); | |
488 | loaddebug(next, 1); | |
489 | loaddebug(next, 2); | |
490 | loaddebug(next, 3); | |
491 | /* no 4 and 5 */ | |
492 | loaddebug(next, 6); | |
493 | loaddebug(next, 7); | |
494 | } | |
495 | ||
529e25f6 EB |
496 | if (test_tsk_thread_flag(prev_p, TIF_NOTSC) ^ |
497 | test_tsk_thread_flag(next_p, TIF_NOTSC)) { | |
498 | /* prev and next are different */ | |
499 | if (test_tsk_thread_flag(next_p, TIF_NOTSC)) | |
500 | hard_disable_TSC(); | |
501 | else | |
502 | hard_enable_TSC(); | |
503 | } | |
504 | ||
d3a4f48d SE |
505 | if (test_tsk_thread_flag(next_p, TIF_IO_BITMAP)) { |
506 | /* | |
507 | * Copy the relevant range of the IO bitmap. | |
508 | * Normally this is 128 bytes or less: | |
509 | */ | |
510 | memcpy(tss->io_bitmap, next->io_bitmap_ptr, | |
511 | max(prev->io_bitmap_max, next->io_bitmap_max)); | |
512 | } else if (test_tsk_thread_flag(prev_p, TIF_IO_BITMAP)) { | |
513 | /* | |
514 | * Clear any possible leftover bits: | |
515 | */ | |
516 | memset(tss->io_bitmap, 0xff, prev->io_bitmap_max); | |
517 | } | |
eee3af4a | 518 | |
b4ef95de | 519 | #ifdef X86_BTS |
eee3af4a MM |
520 | if (test_tsk_thread_flag(prev_p, TIF_BTS_TRACE_TS)) |
521 | ptrace_bts_take_timestamp(prev_p, BTS_TASK_DEPARTS); | |
522 | ||
523 | if (test_tsk_thread_flag(next_p, TIF_BTS_TRACE_TS)) | |
524 | ptrace_bts_take_timestamp(next_p, BTS_TASK_ARRIVES); | |
b4ef95de | 525 | #endif |
d3a4f48d SE |
526 | } |
527 | ||
1da177e4 LT |
528 | /* |
529 | * switch_to(x,y) should switch tasks from x to y. | |
530 | * | |
6612538c | 531 | * This could still be optimized: |
1da177e4 LT |
532 | * - fold all the options into a flag word and test it with a single test. |
533 | * - could test fs/gs bitsliced | |
099f318b AK |
534 | * |
535 | * Kprobes not supported here. Set the probe on schedule instead. | |
1da177e4 | 536 | */ |
f438d914 | 537 | struct task_struct * |
a88cde13 | 538 | __switch_to(struct task_struct *prev_p, struct task_struct *next_p) |
1da177e4 | 539 | { |
87b935a0 JF |
540 | struct thread_struct *prev = &prev_p->thread; |
541 | struct thread_struct *next = &next_p->thread; | |
6612538c | 542 | int cpu = smp_processor_id(); |
1da177e4 | 543 | struct tss_struct *tss = &per_cpu(init_tss, cpu); |
478de5a9 | 544 | unsigned fsindex, gsindex; |
1da177e4 | 545 | |
e07e23e1 AV |
546 | /* we're going to use this soon, after a few expensive things */ |
547 | if (next_p->fpu_counter>5) | |
61c4628b | 548 | prefetch(next->xstate); |
e07e23e1 | 549 | |
1da177e4 LT |
550 | /* |
551 | * Reload esp0, LDT and the page table pointer: | |
552 | */ | |
7818a1e0 | 553 | load_sp0(tss, next); |
1da177e4 LT |
554 | |
555 | /* | |
556 | * Switch DS and ES. | |
557 | * This won't pick up thread selector changes, but I guess that is ok. | |
558 | */ | |
ada85708 | 559 | savesegment(es, prev->es); |
1da177e4 LT |
560 | if (unlikely(next->es | prev->es)) |
561 | loadsegment(es, next->es); | |
ada85708 JF |
562 | |
563 | savesegment(ds, prev->ds); | |
1da177e4 LT |
564 | if (unlikely(next->ds | prev->ds)) |
565 | loadsegment(ds, next->ds); | |
566 | ||
478de5a9 JF |
567 | |
568 | /* We must save %fs and %gs before load_TLS() because | |
569 | * %fs and %gs may be cleared by load_TLS(). | |
570 | * | |
571 | * (e.g. xen_load_tls()) | |
572 | */ | |
573 | savesegment(fs, fsindex); | |
574 | savesegment(gs, gsindex); | |
575 | ||
1da177e4 LT |
576 | load_TLS(next, cpu); |
577 | ||
3fe0a63e JF |
578 | /* |
579 | * Leave lazy mode, flushing any hypercalls made here. | |
580 | * This must be done before restoring TLS segments so | |
581 | * the GDT and LDT are properly updated, and must be | |
582 | * done before math_state_restore, so the TS bit is up | |
583 | * to date. | |
584 | */ | |
585 | arch_leave_lazy_cpu_mode(); | |
586 | ||
1da177e4 LT |
587 | /* |
588 | * Switch FS and GS. | |
87b935a0 JF |
589 | * |
590 | * Segment register != 0 always requires a reload. Also | |
591 | * reload when it has changed. When prev process used 64bit | |
592 | * base always reload to avoid an information leak. | |
1da177e4 | 593 | */ |
87b935a0 JF |
594 | if (unlikely(fsindex | next->fsindex | prev->fs)) { |
595 | loadsegment(fs, next->fsindex); | |
596 | /* | |
597 | * Check if the user used a selector != 0; if yes | |
598 | * clear 64bit base, since overloaded base is always | |
599 | * mapped to the Null selector | |
600 | */ | |
601 | if (fsindex) | |
1da177e4 | 602 | prev->fs = 0; |
87b935a0 JF |
603 | } |
604 | /* when next process has a 64bit base use it */ | |
605 | if (next->fs) | |
606 | wrmsrl(MSR_FS_BASE, next->fs); | |
607 | prev->fsindex = fsindex; | |
608 | ||
609 | if (unlikely(gsindex | next->gsindex | prev->gs)) { | |
610 | load_gs_index(next->gsindex); | |
611 | if (gsindex) | |
1da177e4 | 612 | prev->gs = 0; |
1da177e4 | 613 | } |
87b935a0 JF |
614 | if (next->gs) |
615 | wrmsrl(MSR_KERNEL_GS_BASE, next->gs); | |
616 | prev->gsindex = gsindex; | |
1da177e4 | 617 | |
0a5ace2a AK |
618 | /* Must be after DS reload */ |
619 | unlazy_fpu(prev_p); | |
620 | ||
1da177e4 | 621 | /* |
45948d77 | 622 | * Switch the PDA and FPU contexts. |
1da177e4 | 623 | */ |
faca6227 PA |
624 | prev->usersp = read_pda(oldrsp); |
625 | write_pda(oldrsp, next->usersp); | |
1da177e4 | 626 | write_pda(pcurrent, next_p); |
18bd057b | 627 | |
a88cde13 | 628 | write_pda(kernelstack, |
87b935a0 JF |
629 | (unsigned long)task_stack_page(next_p) + |
630 | THREAD_SIZE - PDA_STACKOFFSET); | |
0a425405 AV |
631 | #ifdef CONFIG_CC_STACKPROTECTOR |
632 | write_pda(stack_canary, next_p->stack_canary); | |
633 | /* | |
634 | * Build time only check to make sure the stack_canary is at | |
635 | * offset 40 in the pda; this is a gcc ABI requirement | |
636 | */ | |
637 | BUILD_BUG_ON(offsetof(struct x8664_pda, stack_canary) != 40); | |
638 | #endif | |
1da177e4 LT |
639 | |
640 | /* | |
d3a4f48d | 641 | * Now maybe reload the debug registers and handle I/O bitmaps |
1da177e4 | 642 | */ |
eee3af4a MM |
643 | if (unlikely(task_thread_info(next_p)->flags & _TIF_WORK_CTXSW_NEXT || |
644 | task_thread_info(prev_p)->flags & _TIF_WORK_CTXSW_PREV)) | |
d3a4f48d | 645 | __switch_to_xtra(prev_p, next_p, tss); |
1da177e4 | 646 | |
e07e23e1 AV |
647 | /* If the task has used fpu the last 5 timeslices, just do a full |
648 | * restore of the math state immediately to avoid the trap; the | |
649 | * chances of needing FPU soon are obviously high now | |
870568b3 SS |
650 | * |
651 | * tsk_used_math() checks prevent calling math_state_restore(), | |
652 | * which can sleep in the case of !tsk_used_math() | |
e07e23e1 | 653 | */ |
870568b3 | 654 | if (tsk_used_math(next_p) && next_p->fpu_counter > 5) |
e07e23e1 | 655 | math_state_restore(); |
1da177e4 LT |
656 | return prev_p; |
657 | } | |
658 | ||
659 | /* | |
660 | * sys_execve() executes a new program. | |
661 | */ | |
6612538c | 662 | asmlinkage |
1da177e4 | 663 | long sys_execve(char __user *name, char __user * __user *argv, |
5d119b2c | 664 | char __user * __user *envp, struct pt_regs *regs) |
1da177e4 LT |
665 | { |
666 | long error; | |
667 | char * filename; | |
668 | ||
669 | filename = getname(name); | |
670 | error = PTR_ERR(filename); | |
5d119b2c | 671 | if (IS_ERR(filename)) |
1da177e4 | 672 | return error; |
5d119b2c | 673 | error = do_execve(filename, argv, envp, regs); |
1da177e4 LT |
674 | putname(filename); |
675 | return error; | |
676 | } | |
677 | ||
678 | void set_personality_64bit(void) | |
679 | { | |
680 | /* inherit personality from parent */ | |
681 | ||
682 | /* Make sure to be in 64bit mode */ | |
6612538c | 683 | clear_thread_flag(TIF_IA32); |
1da177e4 LT |
684 | |
685 | /* TBD: overwrites user setup. Should have two bits. | |
686 | But 64bit processes have always behaved this way, | |
687 | so it's not too bad. The main problem is just that | |
6612538c | 688 | 32bit childs are affected again. */ |
1da177e4 LT |
689 | current->personality &= ~READ_IMPLIES_EXEC; |
690 | } | |
691 | ||
692 | asmlinkage long sys_fork(struct pt_regs *regs) | |
693 | { | |
65ea5b03 | 694 | return do_fork(SIGCHLD, regs->sp, regs, 0, NULL, NULL); |
1da177e4 LT |
695 | } |
696 | ||
a88cde13 AK |
697 | asmlinkage long |
698 | sys_clone(unsigned long clone_flags, unsigned long newsp, | |
699 | void __user *parent_tid, void __user *child_tid, struct pt_regs *regs) | |
1da177e4 LT |
700 | { |
701 | if (!newsp) | |
65ea5b03 | 702 | newsp = regs->sp; |
1da177e4 LT |
703 | return do_fork(clone_flags, newsp, regs, 0, parent_tid, child_tid); |
704 | } | |
705 | ||
706 | /* | |
707 | * This is trivial, and on the face of it looks like it | |
708 | * could equally well be done in user mode. | |
709 | * | |
710 | * Not so, for quite unobvious reasons - register pressure. | |
711 | * In user mode vfork() cannot have a stack frame, and if | |
712 | * done by calling the "clone()" system call directly, you | |
713 | * do not have enough call-clobbered registers to hold all | |
714 | * the information you need. | |
715 | */ | |
716 | asmlinkage long sys_vfork(struct pt_regs *regs) | |
717 | { | |
65ea5b03 | 718 | return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, regs->sp, regs, 0, |
1da177e4 LT |
719 | NULL, NULL); |
720 | } | |
721 | ||
722 | unsigned long get_wchan(struct task_struct *p) | |
723 | { | |
724 | unsigned long stack; | |
65ea5b03 | 725 | u64 fp,ip; |
1da177e4 LT |
726 | int count = 0; |
727 | ||
728 | if (!p || p == current || p->state==TASK_RUNNING) | |
729 | return 0; | |
57eafdc2 | 730 | stack = (unsigned long)task_stack_page(p); |
faca6227 | 731 | if (p->thread.sp < stack || p->thread.sp > stack+THREAD_SIZE) |
1da177e4 | 732 | return 0; |
faca6227 | 733 | fp = *(u64 *)(p->thread.sp); |
1da177e4 | 734 | do { |
a88cde13 AK |
735 | if (fp < (unsigned long)stack || |
736 | fp > (unsigned long)stack+THREAD_SIZE) | |
1da177e4 | 737 | return 0; |
65ea5b03 PA |
738 | ip = *(u64 *)(fp+8); |
739 | if (!in_sched_functions(ip)) | |
740 | return ip; | |
1da177e4 LT |
741 | fp = *(u64 *)fp; |
742 | } while (count++ < 16); | |
743 | return 0; | |
744 | } | |
745 | ||
746 | long do_arch_prctl(struct task_struct *task, int code, unsigned long addr) | |
747 | { | |
748 | int ret = 0; | |
749 | int doit = task == current; | |
750 | int cpu; | |
751 | ||
752 | switch (code) { | |
753 | case ARCH_SET_GS: | |
84929801 | 754 | if (addr >= TASK_SIZE_OF(task)) |
1da177e4 LT |
755 | return -EPERM; |
756 | cpu = get_cpu(); | |
757 | /* handle small bases via the GDT because that's faster to | |
758 | switch. */ | |
759 | if (addr <= 0xffffffff) { | |
760 | set_32bit_tls(task, GS_TLS, addr); | |
761 | if (doit) { | |
762 | load_TLS(&task->thread, cpu); | |
763 | load_gs_index(GS_TLS_SEL); | |
764 | } | |
765 | task->thread.gsindex = GS_TLS_SEL; | |
766 | task->thread.gs = 0; | |
767 | } else { | |
768 | task->thread.gsindex = 0; | |
769 | task->thread.gs = addr; | |
770 | if (doit) { | |
a88cde13 AK |
771 | load_gs_index(0); |
772 | ret = checking_wrmsrl(MSR_KERNEL_GS_BASE, addr); | |
1da177e4 LT |
773 | } |
774 | } | |
775 | put_cpu(); | |
776 | break; | |
777 | case ARCH_SET_FS: | |
778 | /* Not strictly needed for fs, but do it for symmetry | |
779 | with gs */ | |
84929801 | 780 | if (addr >= TASK_SIZE_OF(task)) |
6612538c | 781 | return -EPERM; |
1da177e4 | 782 | cpu = get_cpu(); |
6612538c | 783 | /* handle small bases via the GDT because that's faster to |
1da177e4 | 784 | switch. */ |
6612538c | 785 | if (addr <= 0xffffffff) { |
1da177e4 | 786 | set_32bit_tls(task, FS_TLS, addr); |
6612538c HS |
787 | if (doit) { |
788 | load_TLS(&task->thread, cpu); | |
ada85708 | 789 | loadsegment(fs, FS_TLS_SEL); |
1da177e4 LT |
790 | } |
791 | task->thread.fsindex = FS_TLS_SEL; | |
792 | task->thread.fs = 0; | |
6612538c | 793 | } else { |
1da177e4 LT |
794 | task->thread.fsindex = 0; |
795 | task->thread.fs = addr; | |
796 | if (doit) { | |
797 | /* set the selector to 0 to not confuse | |
798 | __switch_to */ | |
ada85708 | 799 | loadsegment(fs, 0); |
a88cde13 | 800 | ret = checking_wrmsrl(MSR_FS_BASE, addr); |
1da177e4 LT |
801 | } |
802 | } | |
803 | put_cpu(); | |
804 | break; | |
6612538c HS |
805 | case ARCH_GET_FS: { |
806 | unsigned long base; | |
1da177e4 LT |
807 | if (task->thread.fsindex == FS_TLS_SEL) |
808 | base = read_32bit_tls(task, FS_TLS); | |
a88cde13 | 809 | else if (doit) |
1da177e4 | 810 | rdmsrl(MSR_FS_BASE, base); |
a88cde13 | 811 | else |
1da177e4 | 812 | base = task->thread.fs; |
6612538c HS |
813 | ret = put_user(base, (unsigned long __user *)addr); |
814 | break; | |
1da177e4 | 815 | } |
6612538c | 816 | case ARCH_GET_GS: { |
1da177e4 | 817 | unsigned long base; |
97c2803c | 818 | unsigned gsindex; |
1da177e4 LT |
819 | if (task->thread.gsindex == GS_TLS_SEL) |
820 | base = read_32bit_tls(task, GS_TLS); | |
97c2803c | 821 | else if (doit) { |
ada85708 | 822 | savesegment(gs, gsindex); |
97c2803c JB |
823 | if (gsindex) |
824 | rdmsrl(MSR_KERNEL_GS_BASE, base); | |
825 | else | |
826 | base = task->thread.gs; | |
827 | } | |
a88cde13 | 828 | else |
1da177e4 | 829 | base = task->thread.gs; |
6612538c | 830 | ret = put_user(base, (unsigned long __user *)addr); |
1da177e4 LT |
831 | break; |
832 | } | |
833 | ||
834 | default: | |
835 | ret = -EINVAL; | |
836 | break; | |
6612538c | 837 | } |
1da177e4 | 838 | |
6612538c HS |
839 | return ret; |
840 | } | |
1da177e4 LT |
841 | |
842 | long sys_arch_prctl(int code, unsigned long addr) | |
843 | { | |
844 | return do_arch_prctl(current, code, addr); | |
1da177e4 LT |
845 | } |
846 | ||
847 | unsigned long arch_align_stack(unsigned long sp) | |
848 | { | |
c16b63e0 | 849 | if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space) |
1da177e4 LT |
850 | sp -= get_random_int() % 8192; |
851 | return sp & ~0xf; | |
852 | } | |
c1d171a0 JK |
853 | |
854 | unsigned long arch_randomize_brk(struct mm_struct *mm) | |
855 | { | |
856 | unsigned long range_end = mm->brk + 0x02000000; | |
857 | return randomize_range(mm->brk, range_end, 0) ? : mm->brk; | |
858 | } |