Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/arch/x86-64/kernel/process.c | |
3 | * | |
4 | * Copyright (C) 1995 Linus Torvalds | |
5 | * | |
6 | * Pentium III FXSR, SSE support | |
7 | * Gareth Hughes <gareth@valinux.com>, May 2000 | |
8 | * | |
9 | * X86-64 port | |
10 | * Andi Kleen. | |
11 | * | |
12 | * $Id: process.c,v 1.38 2002/01/15 10:08:03 ak Exp $ | |
13 | */ | |
14 | ||
15 | /* | |
16 | * This file handles the architecture-dependent parts of process handling.. | |
17 | */ | |
18 | ||
19 | #include <stdarg.h> | |
20 | ||
21 | #include <linux/errno.h> | |
22 | #include <linux/sched.h> | |
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> | |
29 | #include <linux/module.h> | |
30 | #include <linux/a.out.h> | |
31 | #include <linux/interrupt.h> | |
32 | #include <linux/delay.h> | |
33 | #include <linux/irq.h> | |
34 | #include <linux/ptrace.h> | |
35 | #include <linux/utsname.h> | |
36 | #include <linux/random.h> | |
73649dab | 37 | #include <linux/kprobes.h> |
1da177e4 LT |
38 | |
39 | #include <asm/uaccess.h> | |
40 | #include <asm/pgtable.h> | |
41 | #include <asm/system.h> | |
42 | #include <asm/io.h> | |
43 | #include <asm/processor.h> | |
44 | #include <asm/i387.h> | |
45 | #include <asm/mmu_context.h> | |
46 | #include <asm/pda.h> | |
47 | #include <asm/prctl.h> | |
48 | #include <asm/kdebug.h> | |
49 | #include <asm/desc.h> | |
50 | #include <asm/proto.h> | |
51 | #include <asm/ia32.h> | |
52 | ||
53 | asmlinkage extern void ret_from_fork(void); | |
54 | ||
55 | unsigned long kernel_thread_flags = CLONE_VM | CLONE_UNTRACED; | |
56 | ||
57 | static atomic_t hlt_counter = ATOMIC_INIT(0); | |
58 | ||
59 | unsigned long boot_option_idle_override = 0; | |
60 | EXPORT_SYMBOL(boot_option_idle_override); | |
61 | ||
62 | /* | |
63 | * Powermanagement idle function, if any.. | |
64 | */ | |
65 | void (*pm_idle)(void); | |
66 | static DEFINE_PER_CPU(unsigned int, cpu_idle_state); | |
67 | ||
68 | void disable_hlt(void) | |
69 | { | |
70 | atomic_inc(&hlt_counter); | |
71 | } | |
72 | ||
73 | EXPORT_SYMBOL(disable_hlt); | |
74 | ||
75 | void enable_hlt(void) | |
76 | { | |
77 | atomic_dec(&hlt_counter); | |
78 | } | |
79 | ||
80 | EXPORT_SYMBOL(enable_hlt); | |
81 | ||
82 | /* | |
83 | * We use this if we don't have any better | |
84 | * idle routine.. | |
85 | */ | |
86 | void default_idle(void) | |
87 | { | |
88 | if (!atomic_read(&hlt_counter)) { | |
89 | local_irq_disable(); | |
90 | if (!need_resched()) | |
91 | safe_halt(); | |
92 | else | |
93 | local_irq_enable(); | |
94 | } | |
95 | } | |
96 | ||
97 | /* | |
98 | * On SMP it's slightly faster (but much more power-consuming!) | |
99 | * to poll the ->need_resched flag instead of waiting for the | |
100 | * cross-CPU IPI to arrive. Use this option with caution. | |
101 | */ | |
102 | static void poll_idle (void) | |
103 | { | |
104 | int oldval; | |
105 | ||
106 | local_irq_enable(); | |
107 | ||
108 | /* | |
109 | * Deal with another CPU just having chosen a thread to | |
110 | * run here: | |
111 | */ | |
112 | oldval = test_and_clear_thread_flag(TIF_NEED_RESCHED); | |
113 | ||
114 | if (!oldval) { | |
115 | set_thread_flag(TIF_POLLING_NRFLAG); | |
116 | asm volatile( | |
117 | "2:" | |
118 | "testl %0,%1;" | |
119 | "rep; nop;" | |
120 | "je 2b;" | |
121 | : : | |
122 | "i" (_TIF_NEED_RESCHED), | |
123 | "m" (current_thread_info()->flags)); | |
124 | } else { | |
125 | set_need_resched(); | |
126 | } | |
127 | } | |
128 | ||
129 | void cpu_idle_wait(void) | |
130 | { | |
131 | unsigned int cpu, this_cpu = get_cpu(); | |
132 | cpumask_t map; | |
133 | ||
134 | set_cpus_allowed(current, cpumask_of_cpu(this_cpu)); | |
135 | put_cpu(); | |
136 | ||
137 | cpus_clear(map); | |
138 | for_each_online_cpu(cpu) { | |
139 | per_cpu(cpu_idle_state, cpu) = 1; | |
140 | cpu_set(cpu, map); | |
141 | } | |
142 | ||
143 | __get_cpu_var(cpu_idle_state) = 0; | |
144 | ||
145 | wmb(); | |
146 | do { | |
147 | ssleep(1); | |
148 | for_each_online_cpu(cpu) { | |
149 | if (cpu_isset(cpu, map) && !per_cpu(cpu_idle_state, cpu)) | |
150 | cpu_clear(cpu, map); | |
151 | } | |
152 | cpus_and(map, map, cpu_online_map); | |
153 | } while (!cpus_empty(map)); | |
154 | } | |
155 | EXPORT_SYMBOL_GPL(cpu_idle_wait); | |
156 | ||
157 | /* | |
158 | * The idle thread. There's no useful work to be | |
159 | * done, so just try to conserve power and have a | |
160 | * low exit latency (ie sit in a loop waiting for | |
161 | * somebody to say that they'd like to reschedule) | |
162 | */ | |
163 | void cpu_idle (void) | |
164 | { | |
165 | /* endless idle loop with no priority at all */ | |
166 | while (1) { | |
167 | while (!need_resched()) { | |
168 | void (*idle)(void); | |
169 | ||
170 | if (__get_cpu_var(cpu_idle_state)) | |
171 | __get_cpu_var(cpu_idle_state) = 0; | |
172 | ||
173 | rmb(); | |
174 | idle = pm_idle; | |
175 | if (!idle) | |
176 | idle = default_idle; | |
177 | idle(); | |
178 | } | |
179 | ||
180 | schedule(); | |
181 | } | |
182 | } | |
183 | ||
184 | /* | |
185 | * This uses new MONITOR/MWAIT instructions on P4 processors with PNI, | |
186 | * which can obviate IPI to trigger checking of need_resched. | |
187 | * We execute MONITOR against need_resched and enter optimized wait state | |
188 | * through MWAIT. Whenever someone changes need_resched, we would be woken | |
189 | * up from MWAIT (without an IPI). | |
190 | */ | |
191 | static void mwait_idle(void) | |
192 | { | |
193 | local_irq_enable(); | |
194 | ||
195 | if (!need_resched()) { | |
196 | set_thread_flag(TIF_POLLING_NRFLAG); | |
197 | do { | |
198 | __monitor((void *)¤t_thread_info()->flags, 0, 0); | |
199 | if (need_resched()) | |
200 | break; | |
201 | __mwait(0, 0); | |
202 | } while (!need_resched()); | |
203 | clear_thread_flag(TIF_POLLING_NRFLAG); | |
204 | } | |
205 | } | |
206 | ||
e6982c67 | 207 | void __cpuinit select_idle_routine(const struct cpuinfo_x86 *c) |
1da177e4 LT |
208 | { |
209 | static int printed; | |
210 | if (cpu_has(c, X86_FEATURE_MWAIT)) { | |
211 | /* | |
212 | * Skip, if setup has overridden idle. | |
213 | * One CPU supports mwait => All CPUs supports mwait | |
214 | */ | |
215 | if (!pm_idle) { | |
216 | if (!printed) { | |
217 | printk("using mwait in idle threads.\n"); | |
218 | printed = 1; | |
219 | } | |
220 | pm_idle = mwait_idle; | |
221 | } | |
222 | } | |
223 | } | |
224 | ||
225 | static int __init idle_setup (char *str) | |
226 | { | |
227 | if (!strncmp(str, "poll", 4)) { | |
228 | printk("using polling idle threads.\n"); | |
229 | pm_idle = poll_idle; | |
230 | } | |
231 | ||
232 | boot_option_idle_override = 1; | |
233 | return 1; | |
234 | } | |
235 | ||
236 | __setup("idle=", idle_setup); | |
237 | ||
238 | /* Prints also some state that isn't saved in the pt_regs */ | |
239 | void __show_regs(struct pt_regs * regs) | |
240 | { | |
241 | unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L, fs, gs, shadowgs; | |
242 | unsigned int fsindex,gsindex; | |
243 | unsigned int ds,cs,es; | |
244 | ||
245 | printk("\n"); | |
246 | print_modules(); | |
247 | printk("Pid: %d, comm: %.20s %s %s\n", | |
248 | current->pid, current->comm, print_tainted(), system_utsname.release); | |
249 | printk("RIP: %04lx:[<%016lx>] ", regs->cs & 0xffff, regs->rip); | |
250 | printk_address(regs->rip); | |
251 | printk("\nRSP: %04lx:%016lx EFLAGS: %08lx\n", regs->ss, regs->rsp, regs->eflags); | |
252 | printk("RAX: %016lx RBX: %016lx RCX: %016lx\n", | |
253 | regs->rax, regs->rbx, regs->rcx); | |
254 | printk("RDX: %016lx RSI: %016lx RDI: %016lx\n", | |
255 | regs->rdx, regs->rsi, regs->rdi); | |
256 | printk("RBP: %016lx R08: %016lx R09: %016lx\n", | |
257 | regs->rbp, regs->r8, regs->r9); | |
258 | printk("R10: %016lx R11: %016lx R12: %016lx\n", | |
259 | regs->r10, regs->r11, regs->r12); | |
260 | printk("R13: %016lx R14: %016lx R15: %016lx\n", | |
261 | regs->r13, regs->r14, regs->r15); | |
262 | ||
263 | asm("movl %%ds,%0" : "=r" (ds)); | |
264 | asm("movl %%cs,%0" : "=r" (cs)); | |
265 | asm("movl %%es,%0" : "=r" (es)); | |
266 | asm("movl %%fs,%0" : "=r" (fsindex)); | |
267 | asm("movl %%gs,%0" : "=r" (gsindex)); | |
268 | ||
269 | rdmsrl(MSR_FS_BASE, fs); | |
270 | rdmsrl(MSR_GS_BASE, gs); | |
271 | rdmsrl(MSR_KERNEL_GS_BASE, shadowgs); | |
272 | ||
273 | asm("movq %%cr0, %0": "=r" (cr0)); | |
274 | asm("movq %%cr2, %0": "=r" (cr2)); | |
275 | asm("movq %%cr3, %0": "=r" (cr3)); | |
276 | asm("movq %%cr4, %0": "=r" (cr4)); | |
277 | ||
278 | printk("FS: %016lx(%04x) GS:%016lx(%04x) knlGS:%016lx\n", | |
279 | fs,fsindex,gs,gsindex,shadowgs); | |
280 | printk("CS: %04x DS: %04x ES: %04x CR0: %016lx\n", cs, ds, es, cr0); | |
281 | printk("CR2: %016lx CR3: %016lx CR4: %016lx\n", cr2, cr3, cr4); | |
282 | } | |
283 | ||
284 | void show_regs(struct pt_regs *regs) | |
285 | { | |
286 | __show_regs(regs); | |
287 | show_trace(®s->rsp); | |
288 | } | |
289 | ||
290 | /* | |
291 | * Free current thread data structures etc.. | |
292 | */ | |
293 | void exit_thread(void) | |
294 | { | |
295 | struct task_struct *me = current; | |
296 | struct thread_struct *t = &me->thread; | |
73649dab RL |
297 | |
298 | /* | |
299 | * Remove function-return probe instances associated with this task | |
300 | * and put them back on the free list. Do not insert an exit probe for | |
301 | * this function, it will be disabled by kprobe_flush_task if you do. | |
302 | */ | |
303 | kprobe_flush_task(me); | |
304 | ||
1da177e4 LT |
305 | if (me->thread.io_bitmap_ptr) { |
306 | struct tss_struct *tss = &per_cpu(init_tss, get_cpu()); | |
307 | ||
308 | kfree(t->io_bitmap_ptr); | |
309 | t->io_bitmap_ptr = NULL; | |
310 | /* | |
311 | * Careful, clear this in the TSS too: | |
312 | */ | |
313 | memset(tss->io_bitmap, 0xff, t->io_bitmap_max); | |
314 | t->io_bitmap_max = 0; | |
315 | put_cpu(); | |
316 | } | |
317 | } | |
318 | ||
319 | void flush_thread(void) | |
320 | { | |
321 | struct task_struct *tsk = current; | |
322 | struct thread_info *t = current_thread_info(); | |
323 | ||
73649dab RL |
324 | /* |
325 | * Remove function-return probe instances associated with this task | |
326 | * and put them back on the free list. Do not insert an exit probe for | |
327 | * this function, it will be disabled by kprobe_flush_task if you do. | |
328 | */ | |
329 | kprobe_flush_task(tsk); | |
330 | ||
1da177e4 LT |
331 | if (t->flags & _TIF_ABI_PENDING) |
332 | t->flags ^= (_TIF_ABI_PENDING | _TIF_IA32); | |
333 | ||
334 | tsk->thread.debugreg0 = 0; | |
335 | tsk->thread.debugreg1 = 0; | |
336 | tsk->thread.debugreg2 = 0; | |
337 | tsk->thread.debugreg3 = 0; | |
338 | tsk->thread.debugreg6 = 0; | |
339 | tsk->thread.debugreg7 = 0; | |
340 | memset(tsk->thread.tls_array, 0, sizeof(tsk->thread.tls_array)); | |
341 | /* | |
342 | * Forget coprocessor state.. | |
343 | */ | |
344 | clear_fpu(tsk); | |
345 | clear_used_math(); | |
346 | } | |
347 | ||
348 | void release_thread(struct task_struct *dead_task) | |
349 | { | |
350 | if (dead_task->mm) { | |
351 | if (dead_task->mm->context.size) { | |
352 | printk("WARNING: dead process %8s still has LDT? <%p/%d>\n", | |
353 | dead_task->comm, | |
354 | dead_task->mm->context.ldt, | |
355 | dead_task->mm->context.size); | |
356 | BUG(); | |
357 | } | |
358 | } | |
359 | } | |
360 | ||
361 | static inline void set_32bit_tls(struct task_struct *t, int tls, u32 addr) | |
362 | { | |
363 | struct user_desc ud = { | |
364 | .base_addr = addr, | |
365 | .limit = 0xfffff, | |
366 | .seg_32bit = 1, | |
367 | .limit_in_pages = 1, | |
368 | .useable = 1, | |
369 | }; | |
370 | struct n_desc_struct *desc = (void *)t->thread.tls_array; | |
371 | desc += tls; | |
372 | desc->a = LDT_entry_a(&ud); | |
373 | desc->b = LDT_entry_b(&ud); | |
374 | } | |
375 | ||
376 | static inline u32 read_32bit_tls(struct task_struct *t, int tls) | |
377 | { | |
378 | struct desc_struct *desc = (void *)t->thread.tls_array; | |
379 | desc += tls; | |
380 | return desc->base0 | | |
381 | (((u32)desc->base1) << 16) | | |
382 | (((u32)desc->base2) << 24); | |
383 | } | |
384 | ||
385 | /* | |
386 | * This gets called before we allocate a new thread and copy | |
387 | * the current task into it. | |
388 | */ | |
389 | void prepare_to_copy(struct task_struct *tsk) | |
390 | { | |
391 | unlazy_fpu(tsk); | |
392 | } | |
393 | ||
394 | int copy_thread(int nr, unsigned long clone_flags, unsigned long rsp, | |
395 | unsigned long unused, | |
396 | struct task_struct * p, struct pt_regs * regs) | |
397 | { | |
398 | int err; | |
399 | struct pt_regs * childregs; | |
400 | struct task_struct *me = current; | |
401 | ||
402 | childregs = ((struct pt_regs *) (THREAD_SIZE + (unsigned long) p->thread_info)) - 1; | |
403 | ||
404 | *childregs = *regs; | |
405 | ||
406 | childregs->rax = 0; | |
407 | childregs->rsp = rsp; | |
408 | if (rsp == ~0UL) { | |
409 | childregs->rsp = (unsigned long)childregs; | |
410 | } | |
411 | ||
412 | p->thread.rsp = (unsigned long) childregs; | |
413 | p->thread.rsp0 = (unsigned long) (childregs+1); | |
414 | p->thread.userrsp = me->thread.userrsp; | |
415 | ||
416 | set_ti_thread_flag(p->thread_info, TIF_FORK); | |
417 | ||
418 | p->thread.fs = me->thread.fs; | |
419 | p->thread.gs = me->thread.gs; | |
420 | ||
fd51f666 L |
421 | asm("mov %%gs,%0" : "=m" (p->thread.gsindex)); |
422 | asm("mov %%fs,%0" : "=m" (p->thread.fsindex)); | |
423 | asm("mov %%es,%0" : "=m" (p->thread.es)); | |
424 | asm("mov %%ds,%0" : "=m" (p->thread.ds)); | |
1da177e4 LT |
425 | |
426 | if (unlikely(me->thread.io_bitmap_ptr != NULL)) { | |
427 | p->thread.io_bitmap_ptr = kmalloc(IO_BITMAP_BYTES, GFP_KERNEL); | |
428 | if (!p->thread.io_bitmap_ptr) { | |
429 | p->thread.io_bitmap_max = 0; | |
430 | return -ENOMEM; | |
431 | } | |
432 | memcpy(p->thread.io_bitmap_ptr, me->thread.io_bitmap_ptr, IO_BITMAP_BYTES); | |
433 | } | |
434 | ||
435 | /* | |
436 | * Set a new TLS for the child thread? | |
437 | */ | |
438 | if (clone_flags & CLONE_SETTLS) { | |
439 | #ifdef CONFIG_IA32_EMULATION | |
440 | if (test_thread_flag(TIF_IA32)) | |
441 | err = ia32_child_tls(p, childregs); | |
442 | else | |
443 | #endif | |
444 | err = do_arch_prctl(p, ARCH_SET_FS, childregs->r8); | |
445 | if (err) | |
446 | goto out; | |
447 | } | |
448 | err = 0; | |
449 | out: | |
450 | if (err && p->thread.io_bitmap_ptr) { | |
451 | kfree(p->thread.io_bitmap_ptr); | |
452 | p->thread.io_bitmap_max = 0; | |
453 | } | |
454 | return err; | |
455 | } | |
456 | ||
457 | /* | |
458 | * This special macro can be used to load a debugging register | |
459 | */ | |
460 | #define loaddebug(thread,r) set_debug(thread->debugreg ## r, r) | |
461 | ||
462 | /* | |
463 | * switch_to(x,y) should switch tasks from x to y. | |
464 | * | |
465 | * This could still be optimized: | |
466 | * - fold all the options into a flag word and test it with a single test. | |
467 | * - could test fs/gs bitsliced | |
468 | */ | |
469 | struct task_struct *__switch_to(struct task_struct *prev_p, struct task_struct *next_p) | |
470 | { | |
471 | struct thread_struct *prev = &prev_p->thread, | |
472 | *next = &next_p->thread; | |
473 | int cpu = smp_processor_id(); | |
474 | struct tss_struct *tss = &per_cpu(init_tss, cpu); | |
475 | ||
476 | unlazy_fpu(prev_p); | |
477 | ||
478 | /* | |
479 | * Reload esp0, LDT and the page table pointer: | |
480 | */ | |
481 | tss->rsp0 = next->rsp0; | |
482 | ||
483 | /* | |
484 | * Switch DS and ES. | |
485 | * This won't pick up thread selector changes, but I guess that is ok. | |
486 | */ | |
fd51f666 | 487 | asm volatile("mov %%es,%0" : "=m" (prev->es)); |
1da177e4 LT |
488 | if (unlikely(next->es | prev->es)) |
489 | loadsegment(es, next->es); | |
490 | ||
fd51f666 | 491 | asm volatile ("mov %%ds,%0" : "=m" (prev->ds)); |
1da177e4 LT |
492 | if (unlikely(next->ds | prev->ds)) |
493 | loadsegment(ds, next->ds); | |
494 | ||
495 | load_TLS(next, cpu); | |
496 | ||
497 | /* | |
498 | * Switch FS and GS. | |
499 | */ | |
500 | { | |
501 | unsigned fsindex; | |
502 | asm volatile("movl %%fs,%0" : "=r" (fsindex)); | |
503 | /* segment register != 0 always requires a reload. | |
504 | also reload when it has changed. | |
505 | when prev process used 64bit base always reload | |
506 | to avoid an information leak. */ | |
507 | if (unlikely(fsindex | next->fsindex | prev->fs)) { | |
508 | loadsegment(fs, next->fsindex); | |
509 | /* check if the user used a selector != 0 | |
510 | * if yes clear 64bit base, since overloaded base | |
511 | * is always mapped to the Null selector | |
512 | */ | |
513 | if (fsindex) | |
514 | prev->fs = 0; | |
515 | } | |
516 | /* when next process has a 64bit base use it */ | |
517 | if (next->fs) | |
518 | wrmsrl(MSR_FS_BASE, next->fs); | |
519 | prev->fsindex = fsindex; | |
520 | } | |
521 | { | |
522 | unsigned gsindex; | |
523 | asm volatile("movl %%gs,%0" : "=r" (gsindex)); | |
524 | if (unlikely(gsindex | next->gsindex | prev->gs)) { | |
525 | load_gs_index(next->gsindex); | |
526 | if (gsindex) | |
527 | prev->gs = 0; | |
528 | } | |
529 | if (next->gs) | |
530 | wrmsrl(MSR_KERNEL_GS_BASE, next->gs); | |
531 | prev->gsindex = gsindex; | |
532 | } | |
533 | ||
534 | /* | |
535 | * Switch the PDA context. | |
536 | */ | |
537 | prev->userrsp = read_pda(oldrsp); | |
538 | write_pda(oldrsp, next->userrsp); | |
539 | write_pda(pcurrent, next_p); | |
540 | write_pda(kernelstack, (unsigned long)next_p->thread_info + THREAD_SIZE - PDA_STACKOFFSET); | |
541 | ||
542 | /* | |
543 | * Now maybe reload the debug registers | |
544 | */ | |
545 | if (unlikely(next->debugreg7)) { | |
546 | loaddebug(next, 0); | |
547 | loaddebug(next, 1); | |
548 | loaddebug(next, 2); | |
549 | loaddebug(next, 3); | |
550 | /* no 4 and 5 */ | |
551 | loaddebug(next, 6); | |
552 | loaddebug(next, 7); | |
553 | } | |
554 | ||
555 | ||
556 | /* | |
557 | * Handle the IO bitmap | |
558 | */ | |
559 | if (unlikely(prev->io_bitmap_ptr || next->io_bitmap_ptr)) { | |
560 | if (next->io_bitmap_ptr) | |
561 | /* | |
562 | * Copy the relevant range of the IO bitmap. | |
563 | * Normally this is 128 bytes or less: | |
564 | */ | |
565 | memcpy(tss->io_bitmap, next->io_bitmap_ptr, | |
566 | max(prev->io_bitmap_max, next->io_bitmap_max)); | |
567 | else { | |
568 | /* | |
569 | * Clear any possible leftover bits: | |
570 | */ | |
571 | memset(tss->io_bitmap, 0xff, prev->io_bitmap_max); | |
572 | } | |
573 | } | |
574 | ||
575 | return prev_p; | |
576 | } | |
577 | ||
578 | /* | |
579 | * sys_execve() executes a new program. | |
580 | */ | |
581 | asmlinkage | |
582 | long sys_execve(char __user *name, char __user * __user *argv, | |
583 | char __user * __user *envp, struct pt_regs regs) | |
584 | { | |
585 | long error; | |
586 | char * filename; | |
587 | ||
588 | filename = getname(name); | |
589 | error = PTR_ERR(filename); | |
590 | if (IS_ERR(filename)) | |
591 | return error; | |
592 | error = do_execve(filename, argv, envp, ®s); | |
593 | if (error == 0) { | |
594 | task_lock(current); | |
595 | current->ptrace &= ~PT_DTRACE; | |
596 | task_unlock(current); | |
597 | } | |
598 | putname(filename); | |
599 | return error; | |
600 | } | |
601 | ||
602 | void set_personality_64bit(void) | |
603 | { | |
604 | /* inherit personality from parent */ | |
605 | ||
606 | /* Make sure to be in 64bit mode */ | |
607 | clear_thread_flag(TIF_IA32); | |
608 | ||
609 | /* TBD: overwrites user setup. Should have two bits. | |
610 | But 64bit processes have always behaved this way, | |
611 | so it's not too bad. The main problem is just that | |
612 | 32bit childs are affected again. */ | |
613 | current->personality &= ~READ_IMPLIES_EXEC; | |
614 | } | |
615 | ||
616 | asmlinkage long sys_fork(struct pt_regs *regs) | |
617 | { | |
618 | return do_fork(SIGCHLD, regs->rsp, regs, 0, NULL, NULL); | |
619 | } | |
620 | ||
621 | asmlinkage long sys_clone(unsigned long clone_flags, unsigned long newsp, void __user *parent_tid, void __user *child_tid, struct pt_regs *regs) | |
622 | { | |
623 | if (!newsp) | |
624 | newsp = regs->rsp; | |
625 | return do_fork(clone_flags, newsp, regs, 0, parent_tid, child_tid); | |
626 | } | |
627 | ||
628 | /* | |
629 | * This is trivial, and on the face of it looks like it | |
630 | * could equally well be done in user mode. | |
631 | * | |
632 | * Not so, for quite unobvious reasons - register pressure. | |
633 | * In user mode vfork() cannot have a stack frame, and if | |
634 | * done by calling the "clone()" system call directly, you | |
635 | * do not have enough call-clobbered registers to hold all | |
636 | * the information you need. | |
637 | */ | |
638 | asmlinkage long sys_vfork(struct pt_regs *regs) | |
639 | { | |
640 | return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, regs->rsp, regs, 0, | |
641 | NULL, NULL); | |
642 | } | |
643 | ||
644 | unsigned long get_wchan(struct task_struct *p) | |
645 | { | |
646 | unsigned long stack; | |
647 | u64 fp,rip; | |
648 | int count = 0; | |
649 | ||
650 | if (!p || p == current || p->state==TASK_RUNNING) | |
651 | return 0; | |
652 | stack = (unsigned long)p->thread_info; | |
653 | if (p->thread.rsp < stack || p->thread.rsp > stack+THREAD_SIZE) | |
654 | return 0; | |
655 | fp = *(u64 *)(p->thread.rsp); | |
656 | do { | |
657 | if (fp < (unsigned long)stack || fp > (unsigned long)stack+THREAD_SIZE) | |
658 | return 0; | |
659 | rip = *(u64 *)(fp+8); | |
660 | if (!in_sched_functions(rip)) | |
661 | return rip; | |
662 | fp = *(u64 *)fp; | |
663 | } while (count++ < 16); | |
664 | return 0; | |
665 | } | |
666 | ||
667 | long do_arch_prctl(struct task_struct *task, int code, unsigned long addr) | |
668 | { | |
669 | int ret = 0; | |
670 | int doit = task == current; | |
671 | int cpu; | |
672 | ||
673 | switch (code) { | |
674 | case ARCH_SET_GS: | |
84929801 | 675 | if (addr >= TASK_SIZE_OF(task)) |
1da177e4 LT |
676 | return -EPERM; |
677 | cpu = get_cpu(); | |
678 | /* handle small bases via the GDT because that's faster to | |
679 | switch. */ | |
680 | if (addr <= 0xffffffff) { | |
681 | set_32bit_tls(task, GS_TLS, addr); | |
682 | if (doit) { | |
683 | load_TLS(&task->thread, cpu); | |
684 | load_gs_index(GS_TLS_SEL); | |
685 | } | |
686 | task->thread.gsindex = GS_TLS_SEL; | |
687 | task->thread.gs = 0; | |
688 | } else { | |
689 | task->thread.gsindex = 0; | |
690 | task->thread.gs = addr; | |
691 | if (doit) { | |
692 | load_gs_index(0); | |
693 | ret = checking_wrmsrl(MSR_KERNEL_GS_BASE, addr); | |
694 | } | |
695 | } | |
696 | put_cpu(); | |
697 | break; | |
698 | case ARCH_SET_FS: | |
699 | /* Not strictly needed for fs, but do it for symmetry | |
700 | with gs */ | |
84929801 | 701 | if (addr >= TASK_SIZE_OF(task)) |
1da177e4 LT |
702 | return -EPERM; |
703 | cpu = get_cpu(); | |
704 | /* handle small bases via the GDT because that's faster to | |
705 | switch. */ | |
706 | if (addr <= 0xffffffff) { | |
707 | set_32bit_tls(task, FS_TLS, addr); | |
708 | if (doit) { | |
709 | load_TLS(&task->thread, cpu); | |
710 | asm volatile("movl %0,%%fs" :: "r" (FS_TLS_SEL)); | |
711 | } | |
712 | task->thread.fsindex = FS_TLS_SEL; | |
713 | task->thread.fs = 0; | |
714 | } else { | |
715 | task->thread.fsindex = 0; | |
716 | task->thread.fs = addr; | |
717 | if (doit) { | |
718 | /* set the selector to 0 to not confuse | |
719 | __switch_to */ | |
720 | asm volatile("movl %0,%%fs" :: "r" (0)); | |
721 | ret = checking_wrmsrl(MSR_FS_BASE, addr); | |
722 | } | |
723 | } | |
724 | put_cpu(); | |
725 | break; | |
726 | case ARCH_GET_FS: { | |
727 | unsigned long base; | |
728 | if (task->thread.fsindex == FS_TLS_SEL) | |
729 | base = read_32bit_tls(task, FS_TLS); | |
730 | else if (doit) { | |
731 | rdmsrl(MSR_FS_BASE, base); | |
732 | } else | |
733 | base = task->thread.fs; | |
734 | ret = put_user(base, (unsigned long __user *)addr); | |
735 | break; | |
736 | } | |
737 | case ARCH_GET_GS: { | |
738 | unsigned long base; | |
739 | if (task->thread.gsindex == GS_TLS_SEL) | |
740 | base = read_32bit_tls(task, GS_TLS); | |
741 | else if (doit) { | |
742 | rdmsrl(MSR_KERNEL_GS_BASE, base); | |
743 | } else | |
744 | base = task->thread.gs; | |
745 | ret = put_user(base, (unsigned long __user *)addr); | |
746 | break; | |
747 | } | |
748 | ||
749 | default: | |
750 | ret = -EINVAL; | |
751 | break; | |
752 | } | |
753 | ||
754 | return ret; | |
755 | } | |
756 | ||
757 | long sys_arch_prctl(int code, unsigned long addr) | |
758 | { | |
759 | return do_arch_prctl(current, code, addr); | |
760 | } | |
761 | ||
762 | /* | |
763 | * Capture the user space registers if the task is not running (in user space) | |
764 | */ | |
765 | int dump_task_regs(struct task_struct *tsk, elf_gregset_t *regs) | |
766 | { | |
767 | struct pt_regs *pp, ptregs; | |
768 | ||
769 | pp = (struct pt_regs *)(tsk->thread.rsp0); | |
770 | --pp; | |
771 | ||
772 | ptregs = *pp; | |
773 | ptregs.cs &= 0xffff; | |
774 | ptregs.ss &= 0xffff; | |
775 | ||
776 | elf_core_copy_regs(regs, &ptregs); | |
777 | ||
778 | return 1; | |
779 | } | |
780 | ||
781 | unsigned long arch_align_stack(unsigned long sp) | |
782 | { | |
783 | if (randomize_va_space) | |
784 | sp -= get_random_int() % 8192; | |
785 | return sp & ~0xf; | |
786 | } |