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> |
7de08b4e GP |
40 | #include <linux/uaccess.h> |
41 | #include <linux/io.h> | |
8b96f011 | 42 | #include <linux/ftrace.h> |
1da177e4 | 43 | |
1da177e4 LT |
44 | #include <asm/pgtable.h> |
45 | #include <asm/system.h> | |
1da177e4 LT |
46 | #include <asm/processor.h> |
47 | #include <asm/i387.h> | |
48 | #include <asm/mmu_context.h> | |
49 | #include <asm/pda.h> | |
50 | #include <asm/prctl.h> | |
1da177e4 LT |
51 | #include <asm/desc.h> |
52 | #include <asm/proto.h> | |
53 | #include <asm/ia32.h> | |
95833c83 | 54 | #include <asm/idle.h> |
bbc1f698 | 55 | #include <asm/syscalls.h> |
1da177e4 LT |
56 | |
57 | asmlinkage extern void ret_from_fork(void); | |
58 | ||
59 | unsigned long kernel_thread_flags = CLONE_VM | CLONE_UNTRACED; | |
60 | ||
e041c683 | 61 | static ATOMIC_NOTIFIER_HEAD(idle_notifier); |
95833c83 AK |
62 | |
63 | void idle_notifier_register(struct notifier_block *n) | |
64 | { | |
e041c683 | 65 | atomic_notifier_chain_register(&idle_notifier, n); |
95833c83 | 66 | } |
c7d87d79 VP |
67 | EXPORT_SYMBOL_GPL(idle_notifier_register); |
68 | ||
69 | void idle_notifier_unregister(struct notifier_block *n) | |
70 | { | |
71 | atomic_notifier_chain_unregister(&idle_notifier, n); | |
72 | } | |
73 | EXPORT_SYMBOL_GPL(idle_notifier_unregister); | |
95833c83 | 74 | |
95833c83 AK |
75 | void enter_idle(void) |
76 | { | |
a15da49d | 77 | write_pda(isidle, 1); |
e041c683 | 78 | atomic_notifier_call_chain(&idle_notifier, IDLE_START, NULL); |
95833c83 AK |
79 | } |
80 | ||
81 | static void __exit_idle(void) | |
82 | { | |
9446868b | 83 | if (test_and_clear_bit_pda(0, isidle) == 0) |
a15da49d | 84 | return; |
e041c683 | 85 | atomic_notifier_call_chain(&idle_notifier, IDLE_END, NULL); |
95833c83 AK |
86 | } |
87 | ||
88 | /* Called from interrupts to signify idle end */ | |
89 | void exit_idle(void) | |
90 | { | |
a15da49d AK |
91 | /* idle loop has pid 0 */ |
92 | if (current->pid) | |
95833c83 AK |
93 | return; |
94 | __exit_idle(); | |
95 | } | |
96 | ||
913da64b | 97 | #ifndef CONFIG_SMP |
76e4f660 AR |
98 | static inline void play_dead(void) |
99 | { | |
100 | BUG(); | |
101 | } | |
913da64b | 102 | #endif |
76e4f660 | 103 | |
1da177e4 LT |
104 | /* |
105 | * The idle thread. There's no useful work to be | |
106 | * done, so just try to conserve power and have a | |
107 | * low exit latency (ie sit in a loop waiting for | |
108 | * somebody to say that they'd like to reschedule) | |
109 | */ | |
b10db7f0 | 110 | void cpu_idle(void) |
1da177e4 | 111 | { |
495ab9c0 | 112 | current_thread_info()->status |= TS_POLLING; |
1da177e4 LT |
113 | /* endless idle loop with no priority at all */ |
114 | while (1) { | |
b8f8c3cf | 115 | tick_nohz_stop_sched_tick(1); |
1da177e4 | 116 | while (!need_resched()) { |
1da177e4 | 117 | |
1da177e4 | 118 | rmb(); |
6ddd2a27 | 119 | |
76e4f660 AR |
120 | if (cpu_is_offline(smp_processor_id())) |
121 | play_dead(); | |
d331e739 VP |
122 | /* |
123 | * Idle routines should keep interrupts disabled | |
124 | * from here on, until they go to idle. | |
125 | * Otherwise, idle callbacks can misfire. | |
126 | */ | |
127 | local_irq_disable(); | |
95833c83 | 128 | enter_idle(); |
81d68a96 SR |
129 | /* Don't trace irqs off for idle */ |
130 | stop_critical_timings(); | |
6ddd2a27 | 131 | pm_idle(); |
81d68a96 | 132 | start_critical_timings(); |
a15da49d AK |
133 | /* In many cases the interrupt that ended idle |
134 | has already called exit_idle. But some idle | |
135 | loops can be woken up without interrupt. */ | |
95833c83 | 136 | __exit_idle(); |
1da177e4 LT |
137 | } |
138 | ||
02290683 | 139 | tick_nohz_restart_sched_tick(); |
5bfb5d69 | 140 | preempt_enable_no_resched(); |
1da177e4 | 141 | schedule(); |
5bfb5d69 | 142 | preempt_disable(); |
1da177e4 LT |
143 | } |
144 | } | |
145 | ||
6612538c | 146 | /* Prints also some state that isn't saved in the pt_regs */ |
e2ce07c8 | 147 | void __show_regs(struct pt_regs *regs, int all) |
1da177e4 LT |
148 | { |
149 | unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L, fs, gs, shadowgs; | |
bb1995d5 | 150 | unsigned long d0, d1, d2, d3, d6, d7; |
6612538c HS |
151 | unsigned int fsindex, gsindex; |
152 | unsigned int ds, cs, es; | |
1da177e4 LT |
153 | |
154 | printk("\n"); | |
155 | print_modules(); | |
8092c654 | 156 | printk(KERN_INFO "Pid: %d, comm: %.20s %s %s %.*s\n", |
9acf23c4 | 157 | current->pid, current->comm, print_tainted(), |
96b644bd SH |
158 | init_utsname()->release, |
159 | (int)strcspn(init_utsname()->version, " "), | |
160 | init_utsname()->version); | |
8092c654 | 161 | printk(KERN_INFO "RIP: %04lx:[<%016lx>] ", regs->cs & 0xffff, regs->ip); |
aafbd7eb | 162 | printk_address(regs->ip, 1); |
8092c654 GP |
163 | printk(KERN_INFO "RSP: %04lx:%016lx EFLAGS: %08lx\n", regs->ss, |
164 | regs->sp, regs->flags); | |
165 | printk(KERN_INFO "RAX: %016lx RBX: %016lx RCX: %016lx\n", | |
65ea5b03 | 166 | regs->ax, regs->bx, regs->cx); |
8092c654 | 167 | printk(KERN_INFO "RDX: %016lx RSI: %016lx RDI: %016lx\n", |
65ea5b03 | 168 | regs->dx, regs->si, regs->di); |
8092c654 | 169 | printk(KERN_INFO "RBP: %016lx R08: %016lx R09: %016lx\n", |
65ea5b03 | 170 | regs->bp, regs->r8, regs->r9); |
8092c654 | 171 | printk(KERN_INFO "R10: %016lx R11: %016lx R12: %016lx\n", |
7de08b4e | 172 | regs->r10, regs->r11, regs->r12); |
8092c654 | 173 | printk(KERN_INFO "R13: %016lx R14: %016lx R15: %016lx\n", |
7de08b4e | 174 | regs->r13, regs->r14, regs->r15); |
1da177e4 | 175 | |
7de08b4e GP |
176 | asm("movl %%ds,%0" : "=r" (ds)); |
177 | asm("movl %%cs,%0" : "=r" (cs)); | |
178 | asm("movl %%es,%0" : "=r" (es)); | |
1da177e4 LT |
179 | asm("movl %%fs,%0" : "=r" (fsindex)); |
180 | asm("movl %%gs,%0" : "=r" (gsindex)); | |
181 | ||
182 | rdmsrl(MSR_FS_BASE, fs); | |
7de08b4e GP |
183 | rdmsrl(MSR_GS_BASE, gs); |
184 | rdmsrl(MSR_KERNEL_GS_BASE, shadowgs); | |
1da177e4 | 185 | |
e2ce07c8 PE |
186 | if (!all) |
187 | return; | |
1da177e4 | 188 | |
f51c9452 GOC |
189 | cr0 = read_cr0(); |
190 | cr2 = read_cr2(); | |
191 | cr3 = read_cr3(); | |
192 | cr4 = read_cr4(); | |
1da177e4 | 193 | |
8092c654 | 194 | printk(KERN_INFO "FS: %016lx(%04x) GS:%016lx(%04x) knlGS:%016lx\n", |
7de08b4e | 195 | fs, fsindex, gs, gsindex, shadowgs); |
8092c654 GP |
196 | printk(KERN_INFO "CS: %04x DS: %04x ES: %04x CR0: %016lx\n", cs, ds, |
197 | es, cr0); | |
198 | printk(KERN_INFO "CR2: %016lx CR3: %016lx CR4: %016lx\n", cr2, cr3, | |
199 | cr4); | |
bb1995d5 AS |
200 | |
201 | get_debugreg(d0, 0); | |
202 | get_debugreg(d1, 1); | |
203 | get_debugreg(d2, 2); | |
8092c654 | 204 | printk(KERN_INFO "DR0: %016lx DR1: %016lx DR2: %016lx\n", d0, d1, d2); |
bb1995d5 AS |
205 | get_debugreg(d3, 3); |
206 | get_debugreg(d6, 6); | |
207 | get_debugreg(d7, 7); | |
8092c654 | 208 | printk(KERN_INFO "DR3: %016lx DR6: %016lx DR7: %016lx\n", d3, d6, d7); |
1da177e4 LT |
209 | } |
210 | ||
211 | void show_regs(struct pt_regs *regs) | |
212 | { | |
8092c654 | 213 | printk(KERN_INFO "CPU %d:", smp_processor_id()); |
e2ce07c8 | 214 | __show_regs(regs, 1); |
bc850d6b | 215 | show_trace(NULL, regs, (void *)(regs + 1), regs->bp); |
1da177e4 LT |
216 | } |
217 | ||
218 | /* | |
219 | * Free current thread data structures etc.. | |
220 | */ | |
221 | void exit_thread(void) | |
222 | { | |
223 | struct task_struct *me = current; | |
224 | struct thread_struct *t = &me->thread; | |
73649dab | 225 | |
6612538c | 226 | if (me->thread.io_bitmap_ptr) { |
1da177e4 LT |
227 | struct tss_struct *tss = &per_cpu(init_tss, get_cpu()); |
228 | ||
229 | kfree(t->io_bitmap_ptr); | |
230 | t->io_bitmap_ptr = NULL; | |
d3a4f48d | 231 | clear_thread_flag(TIF_IO_BITMAP); |
1da177e4 LT |
232 | /* |
233 | * Careful, clear this in the TSS too: | |
234 | */ | |
235 | memset(tss->io_bitmap, 0xff, t->io_bitmap_max); | |
236 | t->io_bitmap_max = 0; | |
237 | put_cpu(); | |
238 | } | |
93fa7636 | 239 | #ifdef CONFIG_X86_DS |
c2724775 MM |
240 | /* Free any BTS tracers that have not been properly released. */ |
241 | if (unlikely(current->bts)) { | |
242 | ds_release_bts(current->bts); | |
243 | current->bts = NULL; | |
244 | ||
245 | kfree(current->bts_buffer); | |
246 | current->bts_buffer = NULL; | |
247 | current->bts_size = 0; | |
93fa7636 MM |
248 | } |
249 | #endif /* CONFIG_X86_DS */ | |
1da177e4 LT |
250 | } |
251 | ||
252 | void flush_thread(void) | |
253 | { | |
254 | struct task_struct *tsk = current; | |
1da177e4 | 255 | |
303cd153 MD |
256 | if (test_tsk_thread_flag(tsk, TIF_ABI_PENDING)) { |
257 | clear_tsk_thread_flag(tsk, TIF_ABI_PENDING); | |
258 | if (test_tsk_thread_flag(tsk, TIF_IA32)) { | |
259 | clear_tsk_thread_flag(tsk, TIF_IA32); | |
260 | } else { | |
261 | set_tsk_thread_flag(tsk, TIF_IA32); | |
4d9bc79c | 262 | current_thread_info()->status |= TS_COMPAT; |
303cd153 | 263 | } |
4d9bc79c | 264 | } |
303cd153 | 265 | clear_tsk_thread_flag(tsk, TIF_DEBUG); |
1da177e4 LT |
266 | |
267 | tsk->thread.debugreg0 = 0; | |
268 | tsk->thread.debugreg1 = 0; | |
269 | tsk->thread.debugreg2 = 0; | |
270 | tsk->thread.debugreg3 = 0; | |
271 | tsk->thread.debugreg6 = 0; | |
272 | tsk->thread.debugreg7 = 0; | |
6612538c | 273 | memset(tsk->thread.tls_array, 0, sizeof(tsk->thread.tls_array)); |
1da177e4 LT |
274 | /* |
275 | * Forget coprocessor state.. | |
276 | */ | |
75118a82 | 277 | tsk->fpu_counter = 0; |
1da177e4 LT |
278 | clear_fpu(tsk); |
279 | clear_used_math(); | |
280 | } | |
281 | ||
282 | void release_thread(struct task_struct *dead_task) | |
283 | { | |
284 | if (dead_task->mm) { | |
285 | if (dead_task->mm->context.size) { | |
286 | printk("WARNING: dead process %8s still has LDT? <%p/%d>\n", | |
287 | dead_task->comm, | |
288 | dead_task->mm->context.ldt, | |
289 | dead_task->mm->context.size); | |
290 | BUG(); | |
291 | } | |
292 | } | |
293 | } | |
294 | ||
295 | static inline void set_32bit_tls(struct task_struct *t, int tls, u32 addr) | |
296 | { | |
6612538c | 297 | struct user_desc ud = { |
1da177e4 LT |
298 | .base_addr = addr, |
299 | .limit = 0xfffff, | |
300 | .seg_32bit = 1, | |
301 | .limit_in_pages = 1, | |
302 | .useable = 1, | |
303 | }; | |
ade1af77 | 304 | struct desc_struct *desc = t->thread.tls_array; |
1da177e4 | 305 | desc += tls; |
80fbb69a | 306 | fill_ldt(desc, &ud); |
1da177e4 LT |
307 | } |
308 | ||
309 | static inline u32 read_32bit_tls(struct task_struct *t, int tls) | |
310 | { | |
91394eb0 | 311 | return get_desc_base(&t->thread.tls_array[tls]); |
1da177e4 LT |
312 | } |
313 | ||
314 | /* | |
315 | * This gets called before we allocate a new thread and copy | |
316 | * the current task into it. | |
317 | */ | |
318 | void prepare_to_copy(struct task_struct *tsk) | |
319 | { | |
320 | unlazy_fpu(tsk); | |
321 | } | |
322 | ||
65ea5b03 | 323 | int copy_thread(int nr, unsigned long clone_flags, unsigned long sp, |
1da177e4 | 324 | unsigned long unused, |
7de08b4e | 325 | struct task_struct *p, struct pt_regs *regs) |
1da177e4 LT |
326 | { |
327 | int err; | |
7de08b4e | 328 | struct pt_regs *childregs; |
1da177e4 LT |
329 | struct task_struct *me = current; |
330 | ||
a88cde13 | 331 | childregs = ((struct pt_regs *) |
57eafdc2 | 332 | (THREAD_SIZE + task_stack_page(p))) - 1; |
1da177e4 LT |
333 | *childregs = *regs; |
334 | ||
65ea5b03 PA |
335 | childregs->ax = 0; |
336 | childregs->sp = sp; | |
337 | if (sp == ~0UL) | |
338 | childregs->sp = (unsigned long)childregs; | |
1da177e4 | 339 | |
faca6227 PA |
340 | p->thread.sp = (unsigned long) childregs; |
341 | p->thread.sp0 = (unsigned long) (childregs+1); | |
342 | p->thread.usersp = me->thread.usersp; | |
1da177e4 | 343 | |
e4f17c43 | 344 | set_tsk_thread_flag(p, TIF_FORK); |
1da177e4 LT |
345 | |
346 | p->thread.fs = me->thread.fs; | |
347 | p->thread.gs = me->thread.gs; | |
348 | ||
ada85708 JF |
349 | savesegment(gs, p->thread.gsindex); |
350 | savesegment(fs, p->thread.fsindex); | |
351 | savesegment(es, p->thread.es); | |
352 | savesegment(ds, p->thread.ds); | |
1da177e4 | 353 | |
d3a4f48d | 354 | if (unlikely(test_tsk_thread_flag(me, TIF_IO_BITMAP))) { |
1da177e4 LT |
355 | p->thread.io_bitmap_ptr = kmalloc(IO_BITMAP_BYTES, GFP_KERNEL); |
356 | if (!p->thread.io_bitmap_ptr) { | |
357 | p->thread.io_bitmap_max = 0; | |
358 | return -ENOMEM; | |
359 | } | |
a88cde13 AK |
360 | memcpy(p->thread.io_bitmap_ptr, me->thread.io_bitmap_ptr, |
361 | IO_BITMAP_BYTES); | |
d3a4f48d | 362 | set_tsk_thread_flag(p, TIF_IO_BITMAP); |
6612538c | 363 | } |
1da177e4 LT |
364 | |
365 | /* | |
366 | * Set a new TLS for the child thread? | |
367 | */ | |
368 | if (clone_flags & CLONE_SETTLS) { | |
369 | #ifdef CONFIG_IA32_EMULATION | |
370 | if (test_thread_flag(TIF_IA32)) | |
efd1ca52 | 371 | err = do_set_thread_area(p, -1, |
65ea5b03 | 372 | (struct user_desc __user *)childregs->si, 0); |
7de08b4e GP |
373 | else |
374 | #endif | |
375 | err = do_arch_prctl(p, ARCH_SET_FS, childregs->r8); | |
376 | if (err) | |
1da177e4 LT |
377 | goto out; |
378 | } | |
379 | err = 0; | |
380 | out: | |
381 | if (err && p->thread.io_bitmap_ptr) { | |
382 | kfree(p->thread.io_bitmap_ptr); | |
383 | p->thread.io_bitmap_max = 0; | |
384 | } | |
385 | return err; | |
386 | } | |
387 | ||
513ad84b IM |
388 | void |
389 | start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp) | |
390 | { | |
ada85708 JF |
391 | loadsegment(fs, 0); |
392 | loadsegment(es, 0); | |
393 | loadsegment(ds, 0); | |
513ad84b IM |
394 | load_gs_index(0); |
395 | regs->ip = new_ip; | |
396 | regs->sp = new_sp; | |
397 | write_pda(oldrsp, new_sp); | |
398 | regs->cs = __USER_CS; | |
399 | regs->ss = __USER_DS; | |
400 | regs->flags = 0x200; | |
401 | set_fs(USER_DS); | |
aa283f49 SS |
402 | /* |
403 | * Free the old FP and other extended state | |
404 | */ | |
405 | free_thread_xstate(current); | |
513ad84b IM |
406 | } |
407 | EXPORT_SYMBOL_GPL(start_thread); | |
408 | ||
529e25f6 EB |
409 | static void hard_disable_TSC(void) |
410 | { | |
411 | write_cr4(read_cr4() | X86_CR4_TSD); | |
412 | } | |
413 | ||
414 | void disable_TSC(void) | |
415 | { | |
416 | preempt_disable(); | |
417 | if (!test_and_set_thread_flag(TIF_NOTSC)) | |
418 | /* | |
419 | * Must flip the CPU state synchronously with | |
420 | * TIF_NOTSC in the current running context. | |
421 | */ | |
422 | hard_disable_TSC(); | |
423 | preempt_enable(); | |
424 | } | |
425 | ||
426 | static void hard_enable_TSC(void) | |
427 | { | |
428 | write_cr4(read_cr4() & ~X86_CR4_TSD); | |
429 | } | |
430 | ||
a4928cff | 431 | static void enable_TSC(void) |
529e25f6 EB |
432 | { |
433 | preempt_disable(); | |
434 | if (test_and_clear_thread_flag(TIF_NOTSC)) | |
435 | /* | |
436 | * Must flip the CPU state synchronously with | |
437 | * TIF_NOTSC in the current running context. | |
438 | */ | |
439 | hard_enable_TSC(); | |
440 | preempt_enable(); | |
441 | } | |
442 | ||
443 | int get_tsc_mode(unsigned long adr) | |
444 | { | |
445 | unsigned int val; | |
446 | ||
447 | if (test_thread_flag(TIF_NOTSC)) | |
448 | val = PR_TSC_SIGSEGV; | |
449 | else | |
450 | val = PR_TSC_ENABLE; | |
451 | ||
452 | return put_user(val, (unsigned int __user *)adr); | |
453 | } | |
454 | ||
455 | int set_tsc_mode(unsigned int val) | |
456 | { | |
457 | if (val == PR_TSC_SIGSEGV) | |
458 | disable_TSC(); | |
459 | else if (val == PR_TSC_ENABLE) | |
460 | enable_TSC(); | |
461 | else | |
462 | return -EINVAL; | |
463 | ||
464 | return 0; | |
465 | } | |
466 | ||
1da177e4 LT |
467 | /* |
468 | * This special macro can be used to load a debugging register | |
469 | */ | |
6612538c HS |
470 | #define loaddebug(thread, r) set_debugreg(thread->debugreg ## r, r) |
471 | ||
d3a4f48d | 472 | static inline void __switch_to_xtra(struct task_struct *prev_p, |
6612538c HS |
473 | struct task_struct *next_p, |
474 | struct tss_struct *tss) | |
d3a4f48d SE |
475 | { |
476 | struct thread_struct *prev, *next; | |
477 | ||
478 | prev = &prev_p->thread, | |
479 | next = &next_p->thread; | |
480 | ||
c2724775 MM |
481 | if (test_tsk_thread_flag(next_p, TIF_DS_AREA_MSR) || |
482 | test_tsk_thread_flag(prev_p, TIF_DS_AREA_MSR)) | |
483 | ds_switch_to(prev_p, next_p); | |
484 | else if (next->debugctlmsr != prev->debugctlmsr) | |
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 | } | |
519 | } | |
520 | ||
1da177e4 LT |
521 | /* |
522 | * switch_to(x,y) should switch tasks from x to y. | |
523 | * | |
6612538c | 524 | * This could still be optimized: |
1da177e4 LT |
525 | * - fold all the options into a flag word and test it with a single test. |
526 | * - could test fs/gs bitsliced | |
099f318b AK |
527 | * |
528 | * Kprobes not supported here. Set the probe on schedule instead. | |
8b96f011 | 529 | * Function graph tracer not supported too. |
1da177e4 | 530 | */ |
8b96f011 | 531 | __notrace_funcgraph struct task_struct * |
a88cde13 | 532 | __switch_to(struct task_struct *prev_p, struct task_struct *next_p) |
1da177e4 | 533 | { |
87b935a0 JF |
534 | struct thread_struct *prev = &prev_p->thread; |
535 | struct thread_struct *next = &next_p->thread; | |
6612538c | 536 | int cpu = smp_processor_id(); |
1da177e4 | 537 | struct tss_struct *tss = &per_cpu(init_tss, cpu); |
478de5a9 | 538 | unsigned fsindex, gsindex; |
1da177e4 | 539 | |
e07e23e1 | 540 | /* we're going to use this soon, after a few expensive things */ |
7de08b4e | 541 | if (next_p->fpu_counter > 5) |
61c4628b | 542 | prefetch(next->xstate); |
e07e23e1 | 543 | |
1da177e4 LT |
544 | /* |
545 | * Reload esp0, LDT and the page table pointer: | |
546 | */ | |
7818a1e0 | 547 | load_sp0(tss, next); |
1da177e4 | 548 | |
7de08b4e | 549 | /* |
1da177e4 LT |
550 | * Switch DS and ES. |
551 | * This won't pick up thread selector changes, but I guess that is ok. | |
552 | */ | |
ada85708 | 553 | savesegment(es, prev->es); |
1da177e4 | 554 | if (unlikely(next->es | prev->es)) |
7de08b4e | 555 | loadsegment(es, next->es); |
ada85708 JF |
556 | |
557 | savesegment(ds, prev->ds); | |
1da177e4 LT |
558 | if (unlikely(next->ds | prev->ds)) |
559 | loadsegment(ds, next->ds); | |
560 | ||
478de5a9 JF |
561 | |
562 | /* We must save %fs and %gs before load_TLS() because | |
563 | * %fs and %gs may be cleared by load_TLS(). | |
564 | * | |
565 | * (e.g. xen_load_tls()) | |
566 | */ | |
567 | savesegment(fs, fsindex); | |
568 | savesegment(gs, gsindex); | |
569 | ||
1da177e4 LT |
570 | load_TLS(next, cpu); |
571 | ||
3fe0a63e JF |
572 | /* |
573 | * Leave lazy mode, flushing any hypercalls made here. | |
574 | * This must be done before restoring TLS segments so | |
575 | * the GDT and LDT are properly updated, and must be | |
576 | * done before math_state_restore, so the TS bit is up | |
577 | * to date. | |
578 | */ | |
579 | arch_leave_lazy_cpu_mode(); | |
580 | ||
7de08b4e | 581 | /* |
1da177e4 | 582 | * Switch FS and GS. |
87b935a0 JF |
583 | * |
584 | * Segment register != 0 always requires a reload. Also | |
585 | * reload when it has changed. When prev process used 64bit | |
586 | * base always reload to avoid an information leak. | |
1da177e4 | 587 | */ |
87b935a0 JF |
588 | if (unlikely(fsindex | next->fsindex | prev->fs)) { |
589 | loadsegment(fs, next->fsindex); | |
7de08b4e | 590 | /* |
87b935a0 JF |
591 | * Check if the user used a selector != 0; if yes |
592 | * clear 64bit base, since overloaded base is always | |
593 | * mapped to the Null selector | |
594 | */ | |
595 | if (fsindex) | |
7de08b4e | 596 | prev->fs = 0; |
87b935a0 JF |
597 | } |
598 | /* when next process has a 64bit base use it */ | |
599 | if (next->fs) | |
600 | wrmsrl(MSR_FS_BASE, next->fs); | |
601 | prev->fsindex = fsindex; | |
602 | ||
603 | if (unlikely(gsindex | next->gsindex | prev->gs)) { | |
604 | load_gs_index(next->gsindex); | |
605 | if (gsindex) | |
7de08b4e | 606 | prev->gs = 0; |
1da177e4 | 607 | } |
87b935a0 JF |
608 | if (next->gs) |
609 | wrmsrl(MSR_KERNEL_GS_BASE, next->gs); | |
610 | prev->gsindex = gsindex; | |
1da177e4 | 611 | |
0a5ace2a AK |
612 | /* Must be after DS reload */ |
613 | unlazy_fpu(prev_p); | |
614 | ||
7de08b4e | 615 | /* |
45948d77 | 616 | * Switch the PDA and FPU contexts. |
1da177e4 | 617 | */ |
faca6227 PA |
618 | prev->usersp = read_pda(oldrsp); |
619 | write_pda(oldrsp, next->usersp); | |
7de08b4e | 620 | write_pda(pcurrent, next_p); |
18bd057b | 621 | |
a88cde13 | 622 | write_pda(kernelstack, |
87b935a0 JF |
623 | (unsigned long)task_stack_page(next_p) + |
624 | THREAD_SIZE - PDA_STACKOFFSET); | |
0a425405 AV |
625 | #ifdef CONFIG_CC_STACKPROTECTOR |
626 | write_pda(stack_canary, next_p->stack_canary); | |
627 | /* | |
628 | * Build time only check to make sure the stack_canary is at | |
629 | * offset 40 in the pda; this is a gcc ABI requirement | |
630 | */ | |
631 | BUILD_BUG_ON(offsetof(struct x8664_pda, stack_canary) != 40); | |
632 | #endif | |
1da177e4 LT |
633 | |
634 | /* | |
d3a4f48d | 635 | * Now maybe reload the debug registers and handle I/O bitmaps |
1da177e4 | 636 | */ |
eee3af4a MM |
637 | if (unlikely(task_thread_info(next_p)->flags & _TIF_WORK_CTXSW_NEXT || |
638 | task_thread_info(prev_p)->flags & _TIF_WORK_CTXSW_PREV)) | |
d3a4f48d | 639 | __switch_to_xtra(prev_p, next_p, tss); |
1da177e4 | 640 | |
e07e23e1 AV |
641 | /* If the task has used fpu the last 5 timeslices, just do a full |
642 | * restore of the math state immediately to avoid the trap; the | |
643 | * chances of needing FPU soon are obviously high now | |
870568b3 SS |
644 | * |
645 | * tsk_used_math() checks prevent calling math_state_restore(), | |
646 | * which can sleep in the case of !tsk_used_math() | |
e07e23e1 | 647 | */ |
870568b3 | 648 | if (tsk_used_math(next_p) && next_p->fpu_counter > 5) |
e07e23e1 | 649 | math_state_restore(); |
1da177e4 LT |
650 | return prev_p; |
651 | } | |
652 | ||
653 | /* | |
654 | * sys_execve() executes a new program. | |
655 | */ | |
6612538c | 656 | asmlinkage |
1da177e4 | 657 | long sys_execve(char __user *name, char __user * __user *argv, |
5d119b2c | 658 | char __user * __user *envp, struct pt_regs *regs) |
1da177e4 LT |
659 | { |
660 | long error; | |
7de08b4e | 661 | char *filename; |
1da177e4 LT |
662 | |
663 | filename = getname(name); | |
664 | error = PTR_ERR(filename); | |
5d119b2c | 665 | if (IS_ERR(filename)) |
1da177e4 | 666 | return error; |
5d119b2c | 667 | error = do_execve(filename, argv, envp, regs); |
1da177e4 LT |
668 | putname(filename); |
669 | return error; | |
670 | } | |
671 | ||
672 | void set_personality_64bit(void) | |
673 | { | |
674 | /* inherit personality from parent */ | |
675 | ||
676 | /* Make sure to be in 64bit mode */ | |
6612538c | 677 | clear_thread_flag(TIF_IA32); |
1da177e4 LT |
678 | |
679 | /* TBD: overwrites user setup. Should have two bits. | |
680 | But 64bit processes have always behaved this way, | |
681 | so it's not too bad. The main problem is just that | |
6612538c | 682 | 32bit childs are affected again. */ |
1da177e4 LT |
683 | current->personality &= ~READ_IMPLIES_EXEC; |
684 | } | |
685 | ||
686 | asmlinkage long sys_fork(struct pt_regs *regs) | |
687 | { | |
65ea5b03 | 688 | return do_fork(SIGCHLD, regs->sp, regs, 0, NULL, NULL); |
1da177e4 LT |
689 | } |
690 | ||
a88cde13 AK |
691 | asmlinkage long |
692 | sys_clone(unsigned long clone_flags, unsigned long newsp, | |
693 | void __user *parent_tid, void __user *child_tid, struct pt_regs *regs) | |
1da177e4 LT |
694 | { |
695 | if (!newsp) | |
65ea5b03 | 696 | newsp = regs->sp; |
1da177e4 LT |
697 | return do_fork(clone_flags, newsp, regs, 0, parent_tid, child_tid); |
698 | } | |
699 | ||
700 | /* | |
701 | * This is trivial, and on the face of it looks like it | |
702 | * could equally well be done in user mode. | |
703 | * | |
704 | * Not so, for quite unobvious reasons - register pressure. | |
705 | * In user mode vfork() cannot have a stack frame, and if | |
706 | * done by calling the "clone()" system call directly, you | |
707 | * do not have enough call-clobbered registers to hold all | |
708 | * the information you need. | |
709 | */ | |
710 | asmlinkage long sys_vfork(struct pt_regs *regs) | |
711 | { | |
65ea5b03 | 712 | return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, regs->sp, regs, 0, |
1da177e4 LT |
713 | NULL, NULL); |
714 | } | |
715 | ||
716 | unsigned long get_wchan(struct task_struct *p) | |
717 | { | |
718 | unsigned long stack; | |
7de08b4e | 719 | u64 fp, ip; |
1da177e4 LT |
720 | int count = 0; |
721 | ||
7de08b4e GP |
722 | if (!p || p == current || p->state == TASK_RUNNING) |
723 | return 0; | |
57eafdc2 | 724 | stack = (unsigned long)task_stack_page(p); |
e1e23bb0 | 725 | if (p->thread.sp < stack || p->thread.sp >= stack+THREAD_SIZE) |
1da177e4 | 726 | return 0; |
faca6227 | 727 | fp = *(u64 *)(p->thread.sp); |
7de08b4e | 728 | do { |
a88cde13 | 729 | if (fp < (unsigned long)stack || |
e1e23bb0 | 730 | fp >= (unsigned long)stack+THREAD_SIZE) |
7de08b4e | 731 | return 0; |
65ea5b03 PA |
732 | ip = *(u64 *)(fp+8); |
733 | if (!in_sched_functions(ip)) | |
734 | return ip; | |
7de08b4e GP |
735 | fp = *(u64 *)fp; |
736 | } while (count++ < 16); | |
1da177e4 LT |
737 | return 0; |
738 | } | |
739 | ||
740 | long do_arch_prctl(struct task_struct *task, int code, unsigned long addr) | |
7de08b4e GP |
741 | { |
742 | int ret = 0; | |
1da177e4 LT |
743 | int doit = task == current; |
744 | int cpu; | |
745 | ||
7de08b4e | 746 | switch (code) { |
1da177e4 | 747 | case ARCH_SET_GS: |
84929801 | 748 | if (addr >= TASK_SIZE_OF(task)) |
7de08b4e | 749 | return -EPERM; |
1da177e4 | 750 | cpu = get_cpu(); |
7de08b4e | 751 | /* handle small bases via the GDT because that's faster to |
1da177e4 | 752 | switch. */ |
7de08b4e GP |
753 | if (addr <= 0xffffffff) { |
754 | set_32bit_tls(task, GS_TLS, addr); | |
755 | if (doit) { | |
1da177e4 | 756 | load_TLS(&task->thread, cpu); |
7de08b4e | 757 | load_gs_index(GS_TLS_SEL); |
1da177e4 | 758 | } |
7de08b4e | 759 | task->thread.gsindex = GS_TLS_SEL; |
1da177e4 | 760 | task->thread.gs = 0; |
7de08b4e | 761 | } else { |
1da177e4 LT |
762 | task->thread.gsindex = 0; |
763 | task->thread.gs = addr; | |
764 | if (doit) { | |
a88cde13 AK |
765 | load_gs_index(0); |
766 | ret = checking_wrmsrl(MSR_KERNEL_GS_BASE, addr); | |
7de08b4e | 767 | } |
1da177e4 LT |
768 | } |
769 | put_cpu(); | |
770 | break; | |
771 | case ARCH_SET_FS: | |
772 | /* Not strictly needed for fs, but do it for symmetry | |
773 | with gs */ | |
84929801 | 774 | if (addr >= TASK_SIZE_OF(task)) |
6612538c | 775 | return -EPERM; |
1da177e4 | 776 | cpu = get_cpu(); |
6612538c | 777 | /* handle small bases via the GDT because that's faster to |
1da177e4 | 778 | switch. */ |
6612538c | 779 | if (addr <= 0xffffffff) { |
1da177e4 | 780 | set_32bit_tls(task, FS_TLS, addr); |
6612538c HS |
781 | if (doit) { |
782 | load_TLS(&task->thread, cpu); | |
ada85708 | 783 | loadsegment(fs, FS_TLS_SEL); |
1da177e4 LT |
784 | } |
785 | task->thread.fsindex = FS_TLS_SEL; | |
786 | task->thread.fs = 0; | |
6612538c | 787 | } else { |
1da177e4 LT |
788 | task->thread.fsindex = 0; |
789 | task->thread.fs = addr; | |
790 | if (doit) { | |
791 | /* set the selector to 0 to not confuse | |
792 | __switch_to */ | |
ada85708 | 793 | loadsegment(fs, 0); |
a88cde13 | 794 | ret = checking_wrmsrl(MSR_FS_BASE, addr); |
1da177e4 LT |
795 | } |
796 | } | |
797 | put_cpu(); | |
798 | break; | |
6612538c HS |
799 | case ARCH_GET_FS: { |
800 | unsigned long base; | |
1da177e4 LT |
801 | if (task->thread.fsindex == FS_TLS_SEL) |
802 | base = read_32bit_tls(task, FS_TLS); | |
a88cde13 | 803 | else if (doit) |
1da177e4 | 804 | rdmsrl(MSR_FS_BASE, base); |
a88cde13 | 805 | else |
1da177e4 | 806 | base = task->thread.fs; |
6612538c HS |
807 | ret = put_user(base, (unsigned long __user *)addr); |
808 | break; | |
1da177e4 | 809 | } |
6612538c | 810 | case ARCH_GET_GS: { |
1da177e4 | 811 | unsigned long base; |
97c2803c | 812 | unsigned gsindex; |
1da177e4 LT |
813 | if (task->thread.gsindex == GS_TLS_SEL) |
814 | base = read_32bit_tls(task, GS_TLS); | |
97c2803c | 815 | else if (doit) { |
ada85708 | 816 | savesegment(gs, gsindex); |
97c2803c JB |
817 | if (gsindex) |
818 | rdmsrl(MSR_KERNEL_GS_BASE, base); | |
819 | else | |
820 | base = task->thread.gs; | |
7de08b4e | 821 | } else |
1da177e4 | 822 | base = task->thread.gs; |
6612538c | 823 | ret = put_user(base, (unsigned long __user *)addr); |
1da177e4 LT |
824 | break; |
825 | } | |
826 | ||
827 | default: | |
828 | ret = -EINVAL; | |
829 | break; | |
6612538c | 830 | } |
1da177e4 | 831 | |
6612538c HS |
832 | return ret; |
833 | } | |
1da177e4 LT |
834 | |
835 | long sys_arch_prctl(int code, unsigned long addr) | |
836 | { | |
837 | return do_arch_prctl(current, code, addr); | |
1da177e4 LT |
838 | } |
839 | ||
840 | unsigned long arch_align_stack(unsigned long sp) | |
841 | { | |
c16b63e0 | 842 | if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space) |
1da177e4 LT |
843 | sp -= get_random_int() % 8192; |
844 | return sp & ~0xf; | |
845 | } | |
c1d171a0 JK |
846 | |
847 | unsigned long arch_randomize_brk(struct mm_struct *mm) | |
848 | { | |
849 | unsigned long range_end = mm->brk + 0x02000000; | |
850 | return randomize_range(mm->brk, range_end, 0) ? : mm->brk; | |
851 | } |