]>
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(); | |
4faac97d TG |
96 | c1e_remove_cpu(raw_smp_processor_id()); |
97 | ||
76e4f660 AR |
98 | mb(); |
99 | /* Ack it */ | |
100 | __get_cpu_var(cpu_state) = CPU_DEAD; | |
101 | ||
1fa744e6 | 102 | local_irq_disable(); |
394a1505 ML |
103 | /* mask all interrupts, flush any and all caches, and halt */ |
104 | wbinvd_halt(); | |
76e4f660 AR |
105 | } |
106 | #else | |
107 | static inline void play_dead(void) | |
108 | { | |
109 | BUG(); | |
110 | } | |
111 | #endif /* CONFIG_HOTPLUG_CPU */ | |
112 | ||
1da177e4 LT |
113 | /* |
114 | * The idle thread. There's no useful work to be | |
115 | * done, so just try to conserve power and have a | |
116 | * low exit latency (ie sit in a loop waiting for | |
117 | * somebody to say that they'd like to reschedule) | |
118 | */ | |
b10db7f0 | 119 | void cpu_idle(void) |
1da177e4 | 120 | { |
495ab9c0 | 121 | current_thread_info()->status |= TS_POLLING; |
1da177e4 LT |
122 | /* endless idle loop with no priority at all */ |
123 | while (1) { | |
b8f8c3cf | 124 | tick_nohz_stop_sched_tick(1); |
1da177e4 | 125 | while (!need_resched()) { |
1da177e4 | 126 | |
1da177e4 | 127 | rmb(); |
6ddd2a27 | 128 | |
76e4f660 AR |
129 | if (cpu_is_offline(smp_processor_id())) |
130 | play_dead(); | |
d331e739 VP |
131 | /* |
132 | * Idle routines should keep interrupts disabled | |
133 | * from here on, until they go to idle. | |
134 | * Otherwise, idle callbacks can misfire. | |
135 | */ | |
136 | local_irq_disable(); | |
95833c83 | 137 | enter_idle(); |
81d68a96 SR |
138 | /* Don't trace irqs off for idle */ |
139 | stop_critical_timings(); | |
6ddd2a27 | 140 | pm_idle(); |
81d68a96 | 141 | start_critical_timings(); |
a15da49d AK |
142 | /* In many cases the interrupt that ended idle |
143 | has already called exit_idle. But some idle | |
144 | loops can be woken up without interrupt. */ | |
95833c83 | 145 | __exit_idle(); |
1da177e4 LT |
146 | } |
147 | ||
02290683 | 148 | tick_nohz_restart_sched_tick(); |
5bfb5d69 | 149 | preempt_enable_no_resched(); |
1da177e4 | 150 | schedule(); |
5bfb5d69 | 151 | preempt_disable(); |
1da177e4 LT |
152 | } |
153 | } | |
154 | ||
6612538c | 155 | /* Prints also some state that isn't saved in the pt_regs */ |
1da177e4 LT |
156 | void __show_regs(struct pt_regs * regs) |
157 | { | |
158 | unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L, fs, gs, shadowgs; | |
bb1995d5 | 159 | unsigned long d0, d1, d2, d3, d6, d7; |
6612538c HS |
160 | unsigned int fsindex, gsindex; |
161 | unsigned int ds, cs, es; | |
1da177e4 LT |
162 | |
163 | printk("\n"); | |
164 | print_modules(); | |
9acf23c4 AK |
165 | printk("Pid: %d, comm: %.20s %s %s %.*s\n", |
166 | current->pid, current->comm, print_tainted(), | |
96b644bd SH |
167 | init_utsname()->release, |
168 | (int)strcspn(init_utsname()->version, " "), | |
169 | init_utsname()->version); | |
65ea5b03 | 170 | printk("RIP: %04lx:[<%016lx>] ", regs->cs & 0xffff, regs->ip); |
aafbd7eb | 171 | printk_address(regs->ip, 1); |
65ea5b03 PA |
172 | printk("RSP: %04lx:%016lx EFLAGS: %08lx\n", regs->ss, regs->sp, |
173 | regs->flags); | |
1da177e4 | 174 | printk("RAX: %016lx RBX: %016lx RCX: %016lx\n", |
65ea5b03 | 175 | regs->ax, regs->bx, regs->cx); |
1da177e4 | 176 | printk("RDX: %016lx RSI: %016lx RDI: %016lx\n", |
65ea5b03 | 177 | regs->dx, regs->si, regs->di); |
1da177e4 | 178 | printk("RBP: %016lx R08: %016lx R09: %016lx\n", |
65ea5b03 | 179 | regs->bp, regs->r8, regs->r9); |
1da177e4 LT |
180 | printk("R10: %016lx R11: %016lx R12: %016lx\n", |
181 | regs->r10, regs->r11, regs->r12); | |
182 | printk("R13: %016lx R14: %016lx R15: %016lx\n", | |
183 | regs->r13, regs->r14, regs->r15); | |
184 | ||
185 | asm("movl %%ds,%0" : "=r" (ds)); | |
186 | asm("movl %%cs,%0" : "=r" (cs)); | |
187 | asm("movl %%es,%0" : "=r" (es)); | |
188 | asm("movl %%fs,%0" : "=r" (fsindex)); | |
189 | asm("movl %%gs,%0" : "=r" (gsindex)); | |
190 | ||
191 | rdmsrl(MSR_FS_BASE, fs); | |
192 | rdmsrl(MSR_GS_BASE, gs); | |
193 | rdmsrl(MSR_KERNEL_GS_BASE, shadowgs); | |
194 | ||
f51c9452 GOC |
195 | cr0 = read_cr0(); |
196 | cr2 = read_cr2(); | |
197 | cr3 = read_cr3(); | |
198 | cr4 = read_cr4(); | |
1da177e4 LT |
199 | |
200 | printk("FS: %016lx(%04x) GS:%016lx(%04x) knlGS:%016lx\n", | |
201 | fs,fsindex,gs,gsindex,shadowgs); | |
202 | printk("CS: %04x DS: %04x ES: %04x CR0: %016lx\n", cs, ds, es, cr0); | |
203 | printk("CR2: %016lx CR3: %016lx CR4: %016lx\n", cr2, cr3, cr4); | |
bb1995d5 AS |
204 | |
205 | get_debugreg(d0, 0); | |
206 | get_debugreg(d1, 1); | |
207 | get_debugreg(d2, 2); | |
208 | printk("DR0: %016lx DR1: %016lx DR2: %016lx\n", d0, d1, d2); | |
209 | get_debugreg(d3, 3); | |
210 | get_debugreg(d6, 6); | |
211 | get_debugreg(d7, 7); | |
212 | printk("DR3: %016lx DR6: %016lx DR7: %016lx\n", d3, d6, d7); | |
1da177e4 LT |
213 | } |
214 | ||
215 | void show_regs(struct pt_regs *regs) | |
216 | { | |
c078d326 | 217 | printk("CPU %d:", smp_processor_id()); |
1da177e4 | 218 | __show_regs(regs); |
bc850d6b | 219 | show_trace(NULL, regs, (void *)(regs + 1), regs->bp); |
1da177e4 LT |
220 | } |
221 | ||
222 | /* | |
223 | * Free current thread data structures etc.. | |
224 | */ | |
225 | void exit_thread(void) | |
226 | { | |
227 | struct task_struct *me = current; | |
228 | struct thread_struct *t = &me->thread; | |
73649dab | 229 | |
6612538c | 230 | if (me->thread.io_bitmap_ptr) { |
1da177e4 LT |
231 | struct tss_struct *tss = &per_cpu(init_tss, get_cpu()); |
232 | ||
233 | kfree(t->io_bitmap_ptr); | |
234 | t->io_bitmap_ptr = NULL; | |
d3a4f48d | 235 | clear_thread_flag(TIF_IO_BITMAP); |
1da177e4 LT |
236 | /* |
237 | * Careful, clear this in the TSS too: | |
238 | */ | |
239 | memset(tss->io_bitmap, 0xff, t->io_bitmap_max); | |
240 | t->io_bitmap_max = 0; | |
241 | put_cpu(); | |
242 | } | |
243 | } | |
244 | ||
245 | void flush_thread(void) | |
246 | { | |
247 | struct task_struct *tsk = current; | |
1da177e4 | 248 | |
303cd153 MD |
249 | if (test_tsk_thread_flag(tsk, TIF_ABI_PENDING)) { |
250 | clear_tsk_thread_flag(tsk, TIF_ABI_PENDING); | |
251 | if (test_tsk_thread_flag(tsk, TIF_IA32)) { | |
252 | clear_tsk_thread_flag(tsk, TIF_IA32); | |
253 | } else { | |
254 | set_tsk_thread_flag(tsk, TIF_IA32); | |
4d9bc79c | 255 | current_thread_info()->status |= TS_COMPAT; |
303cd153 | 256 | } |
4d9bc79c | 257 | } |
303cd153 | 258 | clear_tsk_thread_flag(tsk, TIF_DEBUG); |
1da177e4 LT |
259 | |
260 | tsk->thread.debugreg0 = 0; | |
261 | tsk->thread.debugreg1 = 0; | |
262 | tsk->thread.debugreg2 = 0; | |
263 | tsk->thread.debugreg3 = 0; | |
264 | tsk->thread.debugreg6 = 0; | |
265 | tsk->thread.debugreg7 = 0; | |
6612538c | 266 | memset(tsk->thread.tls_array, 0, sizeof(tsk->thread.tls_array)); |
1da177e4 LT |
267 | /* |
268 | * Forget coprocessor state.. | |
269 | */ | |
75118a82 | 270 | tsk->fpu_counter = 0; |
1da177e4 LT |
271 | clear_fpu(tsk); |
272 | clear_used_math(); | |
273 | } | |
274 | ||
275 | void release_thread(struct task_struct *dead_task) | |
276 | { | |
277 | if (dead_task->mm) { | |
278 | if (dead_task->mm->context.size) { | |
279 | printk("WARNING: dead process %8s still has LDT? <%p/%d>\n", | |
280 | dead_task->comm, | |
281 | dead_task->mm->context.ldt, | |
282 | dead_task->mm->context.size); | |
283 | BUG(); | |
284 | } | |
285 | } | |
286 | } | |
287 | ||
288 | static inline void set_32bit_tls(struct task_struct *t, int tls, u32 addr) | |
289 | { | |
6612538c | 290 | struct user_desc ud = { |
1da177e4 LT |
291 | .base_addr = addr, |
292 | .limit = 0xfffff, | |
293 | .seg_32bit = 1, | |
294 | .limit_in_pages = 1, | |
295 | .useable = 1, | |
296 | }; | |
ade1af77 | 297 | struct desc_struct *desc = t->thread.tls_array; |
1da177e4 | 298 | desc += tls; |
80fbb69a | 299 | fill_ldt(desc, &ud); |
1da177e4 LT |
300 | } |
301 | ||
302 | static inline u32 read_32bit_tls(struct task_struct *t, int tls) | |
303 | { | |
91394eb0 | 304 | return get_desc_base(&t->thread.tls_array[tls]); |
1da177e4 LT |
305 | } |
306 | ||
307 | /* | |
308 | * This gets called before we allocate a new thread and copy | |
309 | * the current task into it. | |
310 | */ | |
311 | void prepare_to_copy(struct task_struct *tsk) | |
312 | { | |
313 | unlazy_fpu(tsk); | |
314 | } | |
315 | ||
65ea5b03 | 316 | int copy_thread(int nr, unsigned long clone_flags, unsigned long sp, |
1da177e4 LT |
317 | unsigned long unused, |
318 | struct task_struct * p, struct pt_regs * regs) | |
319 | { | |
320 | int err; | |
321 | struct pt_regs * childregs; | |
322 | struct task_struct *me = current; | |
323 | ||
a88cde13 | 324 | childregs = ((struct pt_regs *) |
57eafdc2 | 325 | (THREAD_SIZE + task_stack_page(p))) - 1; |
1da177e4 LT |
326 | *childregs = *regs; |
327 | ||
65ea5b03 PA |
328 | childregs->ax = 0; |
329 | childregs->sp = sp; | |
330 | if (sp == ~0UL) | |
331 | childregs->sp = (unsigned long)childregs; | |
1da177e4 | 332 | |
faca6227 PA |
333 | p->thread.sp = (unsigned long) childregs; |
334 | p->thread.sp0 = (unsigned long) (childregs+1); | |
335 | p->thread.usersp = me->thread.usersp; | |
1da177e4 | 336 | |
e4f17c43 | 337 | set_tsk_thread_flag(p, TIF_FORK); |
1da177e4 LT |
338 | |
339 | p->thread.fs = me->thread.fs; | |
340 | p->thread.gs = me->thread.gs; | |
341 | ||
ada85708 JF |
342 | savesegment(gs, p->thread.gsindex); |
343 | savesegment(fs, p->thread.fsindex); | |
344 | savesegment(es, p->thread.es); | |
345 | savesegment(ds, p->thread.ds); | |
1da177e4 | 346 | |
d3a4f48d | 347 | if (unlikely(test_tsk_thread_flag(me, TIF_IO_BITMAP))) { |
1da177e4 LT |
348 | p->thread.io_bitmap_ptr = kmalloc(IO_BITMAP_BYTES, GFP_KERNEL); |
349 | if (!p->thread.io_bitmap_ptr) { | |
350 | p->thread.io_bitmap_max = 0; | |
351 | return -ENOMEM; | |
352 | } | |
a88cde13 AK |
353 | memcpy(p->thread.io_bitmap_ptr, me->thread.io_bitmap_ptr, |
354 | IO_BITMAP_BYTES); | |
d3a4f48d | 355 | set_tsk_thread_flag(p, TIF_IO_BITMAP); |
6612538c | 356 | } |
1da177e4 LT |
357 | |
358 | /* | |
359 | * Set a new TLS for the child thread? | |
360 | */ | |
361 | if (clone_flags & CLONE_SETTLS) { | |
362 | #ifdef CONFIG_IA32_EMULATION | |
363 | if (test_thread_flag(TIF_IA32)) | |
efd1ca52 | 364 | err = do_set_thread_area(p, -1, |
65ea5b03 | 365 | (struct user_desc __user *)childregs->si, 0); |
1da177e4 LT |
366 | else |
367 | #endif | |
368 | err = do_arch_prctl(p, ARCH_SET_FS, childregs->r8); | |
369 | if (err) | |
370 | goto out; | |
371 | } | |
372 | err = 0; | |
373 | out: | |
374 | if (err && p->thread.io_bitmap_ptr) { | |
375 | kfree(p->thread.io_bitmap_ptr); | |
376 | p->thread.io_bitmap_max = 0; | |
377 | } | |
378 | return err; | |
379 | } | |
380 | ||
513ad84b IM |
381 | void |
382 | start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp) | |
383 | { | |
ada85708 JF |
384 | loadsegment(fs, 0); |
385 | loadsegment(es, 0); | |
386 | loadsegment(ds, 0); | |
513ad84b IM |
387 | load_gs_index(0); |
388 | regs->ip = new_ip; | |
389 | regs->sp = new_sp; | |
390 | write_pda(oldrsp, new_sp); | |
391 | regs->cs = __USER_CS; | |
392 | regs->ss = __USER_DS; | |
393 | regs->flags = 0x200; | |
394 | set_fs(USER_DS); | |
aa283f49 SS |
395 | /* |
396 | * Free the old FP and other extended state | |
397 | */ | |
398 | free_thread_xstate(current); | |
513ad84b IM |
399 | } |
400 | EXPORT_SYMBOL_GPL(start_thread); | |
401 | ||
529e25f6 EB |
402 | static void hard_disable_TSC(void) |
403 | { | |
404 | write_cr4(read_cr4() | X86_CR4_TSD); | |
405 | } | |
406 | ||
407 | void disable_TSC(void) | |
408 | { | |
409 | preempt_disable(); | |
410 | if (!test_and_set_thread_flag(TIF_NOTSC)) | |
411 | /* | |
412 | * Must flip the CPU state synchronously with | |
413 | * TIF_NOTSC in the current running context. | |
414 | */ | |
415 | hard_disable_TSC(); | |
416 | preempt_enable(); | |
417 | } | |
418 | ||
419 | static void hard_enable_TSC(void) | |
420 | { | |
421 | write_cr4(read_cr4() & ~X86_CR4_TSD); | |
422 | } | |
423 | ||
a4928cff | 424 | static void enable_TSC(void) |
529e25f6 EB |
425 | { |
426 | preempt_disable(); | |
427 | if (test_and_clear_thread_flag(TIF_NOTSC)) | |
428 | /* | |
429 | * Must flip the CPU state synchronously with | |
430 | * TIF_NOTSC in the current running context. | |
431 | */ | |
432 | hard_enable_TSC(); | |
433 | preempt_enable(); | |
434 | } | |
435 | ||
436 | int get_tsc_mode(unsigned long adr) | |
437 | { | |
438 | unsigned int val; | |
439 | ||
440 | if (test_thread_flag(TIF_NOTSC)) | |
441 | val = PR_TSC_SIGSEGV; | |
442 | else | |
443 | val = PR_TSC_ENABLE; | |
444 | ||
445 | return put_user(val, (unsigned int __user *)adr); | |
446 | } | |
447 | ||
448 | int set_tsc_mode(unsigned int val) | |
449 | { | |
450 | if (val == PR_TSC_SIGSEGV) | |
451 | disable_TSC(); | |
452 | else if (val == PR_TSC_ENABLE) | |
453 | enable_TSC(); | |
454 | else | |
455 | return -EINVAL; | |
456 | ||
457 | return 0; | |
458 | } | |
459 | ||
1da177e4 LT |
460 | /* |
461 | * This special macro can be used to load a debugging register | |
462 | */ | |
6612538c HS |
463 | #define loaddebug(thread, r) set_debugreg(thread->debugreg ## r, r) |
464 | ||
d3a4f48d | 465 | static inline void __switch_to_xtra(struct task_struct *prev_p, |
6612538c HS |
466 | struct task_struct *next_p, |
467 | struct tss_struct *tss) | |
d3a4f48d SE |
468 | { |
469 | struct thread_struct *prev, *next; | |
eee3af4a | 470 | unsigned long debugctl; |
d3a4f48d SE |
471 | |
472 | prev = &prev_p->thread, | |
473 | next = &next_p->thread; | |
474 | ||
eee3af4a MM |
475 | debugctl = prev->debugctlmsr; |
476 | if (next->ds_area_msr != prev->ds_area_msr) { | |
477 | /* we clear debugctl to make sure DS | |
478 | * is not in use when we change it */ | |
479 | debugctl = 0; | |
5b0e5084 | 480 | update_debugctlmsr(0); |
eee3af4a MM |
481 | wrmsrl(MSR_IA32_DS_AREA, next->ds_area_msr); |
482 | } | |
483 | ||
484 | if (next->debugctlmsr != debugctl) | |
5b0e5084 | 485 | update_debugctlmsr(next->debugctlmsr); |
7e991604 | 486 | |
d3a4f48d SE |
487 | if (test_tsk_thread_flag(next_p, TIF_DEBUG)) { |
488 | loaddebug(next, 0); | |
489 | loaddebug(next, 1); | |
490 | loaddebug(next, 2); | |
491 | loaddebug(next, 3); | |
492 | /* no 4 and 5 */ | |
493 | loaddebug(next, 6); | |
494 | loaddebug(next, 7); | |
495 | } | |
496 | ||
529e25f6 EB |
497 | if (test_tsk_thread_flag(prev_p, TIF_NOTSC) ^ |
498 | test_tsk_thread_flag(next_p, TIF_NOTSC)) { | |
499 | /* prev and next are different */ | |
500 | if (test_tsk_thread_flag(next_p, TIF_NOTSC)) | |
501 | hard_disable_TSC(); | |
502 | else | |
503 | hard_enable_TSC(); | |
504 | } | |
505 | ||
d3a4f48d SE |
506 | if (test_tsk_thread_flag(next_p, TIF_IO_BITMAP)) { |
507 | /* | |
508 | * Copy the relevant range of the IO bitmap. | |
509 | * Normally this is 128 bytes or less: | |
510 | */ | |
511 | memcpy(tss->io_bitmap, next->io_bitmap_ptr, | |
512 | max(prev->io_bitmap_max, next->io_bitmap_max)); | |
513 | } else if (test_tsk_thread_flag(prev_p, TIF_IO_BITMAP)) { | |
514 | /* | |
515 | * Clear any possible leftover bits: | |
516 | */ | |
517 | memset(tss->io_bitmap, 0xff, prev->io_bitmap_max); | |
518 | } | |
eee3af4a | 519 | |
b4ef95de | 520 | #ifdef X86_BTS |
eee3af4a MM |
521 | if (test_tsk_thread_flag(prev_p, TIF_BTS_TRACE_TS)) |
522 | ptrace_bts_take_timestamp(prev_p, BTS_TASK_DEPARTS); | |
523 | ||
524 | if (test_tsk_thread_flag(next_p, TIF_BTS_TRACE_TS)) | |
525 | ptrace_bts_take_timestamp(next_p, BTS_TASK_ARRIVES); | |
b4ef95de | 526 | #endif |
d3a4f48d SE |
527 | } |
528 | ||
1da177e4 LT |
529 | /* |
530 | * switch_to(x,y) should switch tasks from x to y. | |
531 | * | |
6612538c | 532 | * This could still be optimized: |
1da177e4 LT |
533 | * - fold all the options into a flag word and test it with a single test. |
534 | * - could test fs/gs bitsliced | |
099f318b AK |
535 | * |
536 | * Kprobes not supported here. Set the probe on schedule instead. | |
1da177e4 | 537 | */ |
f438d914 | 538 | struct task_struct * |
a88cde13 | 539 | __switch_to(struct task_struct *prev_p, struct task_struct *next_p) |
1da177e4 | 540 | { |
87b935a0 JF |
541 | struct thread_struct *prev = &prev_p->thread; |
542 | struct thread_struct *next = &next_p->thread; | |
6612538c | 543 | int cpu = smp_processor_id(); |
1da177e4 | 544 | struct tss_struct *tss = &per_cpu(init_tss, cpu); |
478de5a9 | 545 | unsigned fsindex, gsindex; |
1da177e4 | 546 | |
e07e23e1 AV |
547 | /* we're going to use this soon, after a few expensive things */ |
548 | if (next_p->fpu_counter>5) | |
61c4628b | 549 | prefetch(next->xstate); |
e07e23e1 | 550 | |
1da177e4 LT |
551 | /* |
552 | * Reload esp0, LDT and the page table pointer: | |
553 | */ | |
7818a1e0 | 554 | load_sp0(tss, next); |
1da177e4 LT |
555 | |
556 | /* | |
557 | * Switch DS and ES. | |
558 | * This won't pick up thread selector changes, but I guess that is ok. | |
559 | */ | |
ada85708 | 560 | savesegment(es, prev->es); |
1da177e4 LT |
561 | if (unlikely(next->es | prev->es)) |
562 | loadsegment(es, next->es); | |
ada85708 JF |
563 | |
564 | savesegment(ds, prev->ds); | |
1da177e4 LT |
565 | if (unlikely(next->ds | prev->ds)) |
566 | loadsegment(ds, next->ds); | |
567 | ||
478de5a9 JF |
568 | |
569 | /* We must save %fs and %gs before load_TLS() because | |
570 | * %fs and %gs may be cleared by load_TLS(). | |
571 | * | |
572 | * (e.g. xen_load_tls()) | |
573 | */ | |
574 | savesegment(fs, fsindex); | |
575 | savesegment(gs, gsindex); | |
576 | ||
1da177e4 LT |
577 | load_TLS(next, cpu); |
578 | ||
3fe0a63e JF |
579 | /* |
580 | * Leave lazy mode, flushing any hypercalls made here. | |
581 | * This must be done before restoring TLS segments so | |
582 | * the GDT and LDT are properly updated, and must be | |
583 | * done before math_state_restore, so the TS bit is up | |
584 | * to date. | |
585 | */ | |
586 | arch_leave_lazy_cpu_mode(); | |
587 | ||
1da177e4 LT |
588 | /* |
589 | * Switch FS and GS. | |
87b935a0 JF |
590 | * |
591 | * Segment register != 0 always requires a reload. Also | |
592 | * reload when it has changed. When prev process used 64bit | |
593 | * base always reload to avoid an information leak. | |
1da177e4 | 594 | */ |
87b935a0 JF |
595 | if (unlikely(fsindex | next->fsindex | prev->fs)) { |
596 | loadsegment(fs, next->fsindex); | |
597 | /* | |
598 | * Check if the user used a selector != 0; if yes | |
599 | * clear 64bit base, since overloaded base is always | |
600 | * mapped to the Null selector | |
601 | */ | |
602 | if (fsindex) | |
1da177e4 | 603 | prev->fs = 0; |
87b935a0 JF |
604 | } |
605 | /* when next process has a 64bit base use it */ | |
606 | if (next->fs) | |
607 | wrmsrl(MSR_FS_BASE, next->fs); | |
608 | prev->fsindex = fsindex; | |
609 | ||
610 | if (unlikely(gsindex | next->gsindex | prev->gs)) { | |
611 | load_gs_index(next->gsindex); | |
612 | if (gsindex) | |
1da177e4 | 613 | prev->gs = 0; |
1da177e4 | 614 | } |
87b935a0 JF |
615 | if (next->gs) |
616 | wrmsrl(MSR_KERNEL_GS_BASE, next->gs); | |
617 | prev->gsindex = gsindex; | |
1da177e4 | 618 | |
0a5ace2a AK |
619 | /* Must be after DS reload */ |
620 | unlazy_fpu(prev_p); | |
621 | ||
1da177e4 | 622 | /* |
45948d77 | 623 | * Switch the PDA and FPU contexts. |
1da177e4 | 624 | */ |
faca6227 PA |
625 | prev->usersp = read_pda(oldrsp); |
626 | write_pda(oldrsp, next->usersp); | |
1da177e4 | 627 | write_pda(pcurrent, next_p); |
18bd057b | 628 | |
a88cde13 | 629 | write_pda(kernelstack, |
87b935a0 JF |
630 | (unsigned long)task_stack_page(next_p) + |
631 | THREAD_SIZE - PDA_STACKOFFSET); | |
0a425405 AV |
632 | #ifdef CONFIG_CC_STACKPROTECTOR |
633 | write_pda(stack_canary, next_p->stack_canary); | |
634 | /* | |
635 | * Build time only check to make sure the stack_canary is at | |
636 | * offset 40 in the pda; this is a gcc ABI requirement | |
637 | */ | |
638 | BUILD_BUG_ON(offsetof(struct x8664_pda, stack_canary) != 40); | |
639 | #endif | |
1da177e4 LT |
640 | |
641 | /* | |
d3a4f48d | 642 | * Now maybe reload the debug registers and handle I/O bitmaps |
1da177e4 | 643 | */ |
eee3af4a MM |
644 | if (unlikely(task_thread_info(next_p)->flags & _TIF_WORK_CTXSW_NEXT || |
645 | task_thread_info(prev_p)->flags & _TIF_WORK_CTXSW_PREV)) | |
d3a4f48d | 646 | __switch_to_xtra(prev_p, next_p, tss); |
1da177e4 | 647 | |
e07e23e1 AV |
648 | /* If the task has used fpu the last 5 timeslices, just do a full |
649 | * restore of the math state immediately to avoid the trap; the | |
650 | * chances of needing FPU soon are obviously high now | |
870568b3 SS |
651 | * |
652 | * tsk_used_math() checks prevent calling math_state_restore(), | |
653 | * which can sleep in the case of !tsk_used_math() | |
e07e23e1 | 654 | */ |
870568b3 | 655 | if (tsk_used_math(next_p) && next_p->fpu_counter > 5) |
e07e23e1 | 656 | math_state_restore(); |
1da177e4 LT |
657 | return prev_p; |
658 | } | |
659 | ||
660 | /* | |
661 | * sys_execve() executes a new program. | |
662 | */ | |
6612538c | 663 | asmlinkage |
1da177e4 | 664 | long sys_execve(char __user *name, char __user * __user *argv, |
5d119b2c | 665 | char __user * __user *envp, struct pt_regs *regs) |
1da177e4 LT |
666 | { |
667 | long error; | |
668 | char * filename; | |
669 | ||
670 | filename = getname(name); | |
671 | error = PTR_ERR(filename); | |
5d119b2c | 672 | if (IS_ERR(filename)) |
1da177e4 | 673 | return error; |
5d119b2c | 674 | error = do_execve(filename, argv, envp, regs); |
1da177e4 LT |
675 | putname(filename); |
676 | return error; | |
677 | } | |
678 | ||
679 | void set_personality_64bit(void) | |
680 | { | |
681 | /* inherit personality from parent */ | |
682 | ||
683 | /* Make sure to be in 64bit mode */ | |
6612538c | 684 | clear_thread_flag(TIF_IA32); |
1da177e4 LT |
685 | |
686 | /* TBD: overwrites user setup. Should have two bits. | |
687 | But 64bit processes have always behaved this way, | |
688 | so it's not too bad. The main problem is just that | |
6612538c | 689 | 32bit childs are affected again. */ |
1da177e4 LT |
690 | current->personality &= ~READ_IMPLIES_EXEC; |
691 | } | |
692 | ||
693 | asmlinkage long sys_fork(struct pt_regs *regs) | |
694 | { | |
65ea5b03 | 695 | return do_fork(SIGCHLD, regs->sp, regs, 0, NULL, NULL); |
1da177e4 LT |
696 | } |
697 | ||
a88cde13 AK |
698 | asmlinkage long |
699 | sys_clone(unsigned long clone_flags, unsigned long newsp, | |
700 | void __user *parent_tid, void __user *child_tid, struct pt_regs *regs) | |
1da177e4 LT |
701 | { |
702 | if (!newsp) | |
65ea5b03 | 703 | newsp = regs->sp; |
1da177e4 LT |
704 | return do_fork(clone_flags, newsp, regs, 0, parent_tid, child_tid); |
705 | } | |
706 | ||
707 | /* | |
708 | * This is trivial, and on the face of it looks like it | |
709 | * could equally well be done in user mode. | |
710 | * | |
711 | * Not so, for quite unobvious reasons - register pressure. | |
712 | * In user mode vfork() cannot have a stack frame, and if | |
713 | * done by calling the "clone()" system call directly, you | |
714 | * do not have enough call-clobbered registers to hold all | |
715 | * the information you need. | |
716 | */ | |
717 | asmlinkage long sys_vfork(struct pt_regs *regs) | |
718 | { | |
65ea5b03 | 719 | return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, regs->sp, regs, 0, |
1da177e4 LT |
720 | NULL, NULL); |
721 | } | |
722 | ||
723 | unsigned long get_wchan(struct task_struct *p) | |
724 | { | |
725 | unsigned long stack; | |
65ea5b03 | 726 | u64 fp,ip; |
1da177e4 LT |
727 | int count = 0; |
728 | ||
729 | if (!p || p == current || p->state==TASK_RUNNING) | |
730 | return 0; | |
57eafdc2 | 731 | stack = (unsigned long)task_stack_page(p); |
faca6227 | 732 | if (p->thread.sp < stack || p->thread.sp > stack+THREAD_SIZE) |
1da177e4 | 733 | return 0; |
faca6227 | 734 | fp = *(u64 *)(p->thread.sp); |
1da177e4 | 735 | do { |
a88cde13 AK |
736 | if (fp < (unsigned long)stack || |
737 | fp > (unsigned long)stack+THREAD_SIZE) | |
1da177e4 | 738 | return 0; |
65ea5b03 PA |
739 | ip = *(u64 *)(fp+8); |
740 | if (!in_sched_functions(ip)) | |
741 | return ip; | |
1da177e4 LT |
742 | fp = *(u64 *)fp; |
743 | } while (count++ < 16); | |
744 | return 0; | |
745 | } | |
746 | ||
747 | long do_arch_prctl(struct task_struct *task, int code, unsigned long addr) | |
748 | { | |
749 | int ret = 0; | |
750 | int doit = task == current; | |
751 | int cpu; | |
752 | ||
753 | switch (code) { | |
754 | case ARCH_SET_GS: | |
84929801 | 755 | if (addr >= TASK_SIZE_OF(task)) |
1da177e4 LT |
756 | return -EPERM; |
757 | cpu = get_cpu(); | |
758 | /* handle small bases via the GDT because that's faster to | |
759 | switch. */ | |
760 | if (addr <= 0xffffffff) { | |
761 | set_32bit_tls(task, GS_TLS, addr); | |
762 | if (doit) { | |
763 | load_TLS(&task->thread, cpu); | |
764 | load_gs_index(GS_TLS_SEL); | |
765 | } | |
766 | task->thread.gsindex = GS_TLS_SEL; | |
767 | task->thread.gs = 0; | |
768 | } else { | |
769 | task->thread.gsindex = 0; | |
770 | task->thread.gs = addr; | |
771 | if (doit) { | |
a88cde13 AK |
772 | load_gs_index(0); |
773 | ret = checking_wrmsrl(MSR_KERNEL_GS_BASE, addr); | |
1da177e4 LT |
774 | } |
775 | } | |
776 | put_cpu(); | |
777 | break; | |
778 | case ARCH_SET_FS: | |
779 | /* Not strictly needed for fs, but do it for symmetry | |
780 | with gs */ | |
84929801 | 781 | if (addr >= TASK_SIZE_OF(task)) |
6612538c | 782 | return -EPERM; |
1da177e4 | 783 | cpu = get_cpu(); |
6612538c | 784 | /* handle small bases via the GDT because that's faster to |
1da177e4 | 785 | switch. */ |
6612538c | 786 | if (addr <= 0xffffffff) { |
1da177e4 | 787 | set_32bit_tls(task, FS_TLS, addr); |
6612538c HS |
788 | if (doit) { |
789 | load_TLS(&task->thread, cpu); | |
ada85708 | 790 | loadsegment(fs, FS_TLS_SEL); |
1da177e4 LT |
791 | } |
792 | task->thread.fsindex = FS_TLS_SEL; | |
793 | task->thread.fs = 0; | |
6612538c | 794 | } else { |
1da177e4 LT |
795 | task->thread.fsindex = 0; |
796 | task->thread.fs = addr; | |
797 | if (doit) { | |
798 | /* set the selector to 0 to not confuse | |
799 | __switch_to */ | |
ada85708 | 800 | loadsegment(fs, 0); |
a88cde13 | 801 | ret = checking_wrmsrl(MSR_FS_BASE, addr); |
1da177e4 LT |
802 | } |
803 | } | |
804 | put_cpu(); | |
805 | break; | |
6612538c HS |
806 | case ARCH_GET_FS: { |
807 | unsigned long base; | |
1da177e4 LT |
808 | if (task->thread.fsindex == FS_TLS_SEL) |
809 | base = read_32bit_tls(task, FS_TLS); | |
a88cde13 | 810 | else if (doit) |
1da177e4 | 811 | rdmsrl(MSR_FS_BASE, base); |
a88cde13 | 812 | else |
1da177e4 | 813 | base = task->thread.fs; |
6612538c HS |
814 | ret = put_user(base, (unsigned long __user *)addr); |
815 | break; | |
1da177e4 | 816 | } |
6612538c | 817 | case ARCH_GET_GS: { |
1da177e4 | 818 | unsigned long base; |
97c2803c | 819 | unsigned gsindex; |
1da177e4 LT |
820 | if (task->thread.gsindex == GS_TLS_SEL) |
821 | base = read_32bit_tls(task, GS_TLS); | |
97c2803c | 822 | else if (doit) { |
ada85708 | 823 | savesegment(gs, gsindex); |
97c2803c JB |
824 | if (gsindex) |
825 | rdmsrl(MSR_KERNEL_GS_BASE, base); | |
826 | else | |
827 | base = task->thread.gs; | |
828 | } | |
a88cde13 | 829 | else |
1da177e4 | 830 | base = task->thread.gs; |
6612538c | 831 | ret = put_user(base, (unsigned long __user *)addr); |
1da177e4 LT |
832 | break; |
833 | } | |
834 | ||
835 | default: | |
836 | ret = -EINVAL; | |
837 | break; | |
6612538c | 838 | } |
1da177e4 | 839 | |
6612538c HS |
840 | return ret; |
841 | } | |
1da177e4 LT |
842 | |
843 | long sys_arch_prctl(int code, unsigned long addr) | |
844 | { | |
845 | return do_arch_prctl(current, code, addr); | |
1da177e4 LT |
846 | } |
847 | ||
848 | unsigned long arch_align_stack(unsigned long sp) | |
849 | { | |
c16b63e0 | 850 | if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space) |
1da177e4 LT |
851 | sp -= get_random_int() % 8192; |
852 | return sp & ~0xf; | |
853 | } | |
c1d171a0 JK |
854 | |
855 | unsigned long arch_randomize_brk(struct mm_struct *mm) | |
856 | { | |
857 | unsigned long range_end = mm->brk + 0x02000000; | |
858 | return randomize_range(mm->brk, range_end, 0) ? : mm->brk; | |
859 | } |