]>
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 | ||
42059429 | 19 | #include <linux/stackprotector.h> |
76e4f660 | 20 | #include <linux/cpu.h> |
1da177e4 LT |
21 | #include <linux/errno.h> |
22 | #include <linux/sched.h> | |
6612538c | 23 | #include <linux/fs.h> |
1da177e4 LT |
24 | #include <linux/kernel.h> |
25 | #include <linux/mm.h> | |
26 | #include <linux/elfcore.h> | |
27 | #include <linux/smp.h> | |
28 | #include <linux/slab.h> | |
29 | #include <linux/user.h> | |
1da177e4 | 30 | #include <linux/interrupt.h> |
6612538c | 31 | #include <linux/utsname.h> |
1da177e4 | 32 | #include <linux/delay.h> |
6612538c | 33 | #include <linux/module.h> |
1da177e4 | 34 | #include <linux/ptrace.h> |
1da177e4 | 35 | #include <linux/random.h> |
95833c83 | 36 | #include <linux/notifier.h> |
c6fd91f0 | 37 | #include <linux/kprobes.h> |
1eeb66a1 | 38 | #include <linux/kdebug.h> |
02290683 | 39 | #include <linux/tick.h> |
529e25f6 | 40 | #include <linux/prctl.h> |
7de08b4e GP |
41 | #include <linux/uaccess.h> |
42 | #include <linux/io.h> | |
8b96f011 | 43 | #include <linux/ftrace.h> |
1da177e4 | 44 | |
1da177e4 LT |
45 | #include <asm/pgtable.h> |
46 | #include <asm/system.h> | |
1da177e4 LT |
47 | #include <asm/processor.h> |
48 | #include <asm/i387.h> | |
49 | #include <asm/mmu_context.h> | |
1da177e4 | 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> |
bf53de90 | 56 | #include <asm/ds.h> |
1da177e4 LT |
57 | |
58 | asmlinkage extern void ret_from_fork(void); | |
59 | ||
c6f5e0ac BG |
60 | DEFINE_PER_CPU(struct task_struct *, current_task) = &init_task; |
61 | EXPORT_PER_CPU_SYMBOL(current_task); | |
62 | ||
3d1e42a7 | 63 | DEFINE_PER_CPU(unsigned long, old_rsp); |
c2558e0e | 64 | static DEFINE_PER_CPU(unsigned char, is_idle); |
3d1e42a7 | 65 | |
1da177e4 LT |
66 | unsigned long kernel_thread_flags = CLONE_VM | CLONE_UNTRACED; |
67 | ||
e041c683 | 68 | static ATOMIC_NOTIFIER_HEAD(idle_notifier); |
95833c83 AK |
69 | |
70 | void idle_notifier_register(struct notifier_block *n) | |
71 | { | |
e041c683 | 72 | atomic_notifier_chain_register(&idle_notifier, n); |
95833c83 | 73 | } |
c7d87d79 VP |
74 | EXPORT_SYMBOL_GPL(idle_notifier_register); |
75 | ||
76 | void idle_notifier_unregister(struct notifier_block *n) | |
77 | { | |
78 | atomic_notifier_chain_unregister(&idle_notifier, n); | |
79 | } | |
80 | EXPORT_SYMBOL_GPL(idle_notifier_unregister); | |
95833c83 | 81 | |
95833c83 AK |
82 | void enter_idle(void) |
83 | { | |
c2558e0e | 84 | percpu_write(is_idle, 1); |
e041c683 | 85 | atomic_notifier_call_chain(&idle_notifier, IDLE_START, NULL); |
95833c83 AK |
86 | } |
87 | ||
88 | static void __exit_idle(void) | |
89 | { | |
c2558e0e | 90 | if (x86_test_and_clear_bit_percpu(0, is_idle) == 0) |
a15da49d | 91 | return; |
e041c683 | 92 | atomic_notifier_call_chain(&idle_notifier, IDLE_END, NULL); |
95833c83 AK |
93 | } |
94 | ||
95 | /* Called from interrupts to signify idle end */ | |
96 | void exit_idle(void) | |
97 | { | |
a15da49d AK |
98 | /* idle loop has pid 0 */ |
99 | if (current->pid) | |
95833c83 AK |
100 | return; |
101 | __exit_idle(); | |
102 | } | |
103 | ||
913da64b | 104 | #ifndef CONFIG_SMP |
76e4f660 AR |
105 | static inline void play_dead(void) |
106 | { | |
107 | BUG(); | |
108 | } | |
913da64b | 109 | #endif |
76e4f660 | 110 | |
1da177e4 LT |
111 | /* |
112 | * The idle thread. There's no useful work to be | |
113 | * done, so just try to conserve power and have a | |
114 | * low exit latency (ie sit in a loop waiting for | |
115 | * somebody to say that they'd like to reschedule) | |
116 | */ | |
b10db7f0 | 117 | void cpu_idle(void) |
1da177e4 | 118 | { |
495ab9c0 | 119 | current_thread_info()->status |= TS_POLLING; |
ce22bd92 | 120 | |
ce22bd92 AV |
121 | /* |
122 | * If we're the non-boot CPU, nothing set the PDA stack | |
7e09b2a0 IM |
123 | * canary up for us - and if we are the boot CPU we have |
124 | * a 0 stack canary. This is a good place for updating | |
125 | * it, as we wont ever return from this function (so the | |
126 | * invalid canaries already on the stack wont ever | |
127 | * trigger): | |
ce22bd92 | 128 | */ |
18aa8bb1 IM |
129 | boot_init_stack_canary(); |
130 | ||
1da177e4 LT |
131 | /* endless idle loop with no priority at all */ |
132 | while (1) { | |
b8f8c3cf | 133 | tick_nohz_stop_sched_tick(1); |
1da177e4 | 134 | while (!need_resched()) { |
1da177e4 | 135 | |
1da177e4 | 136 | rmb(); |
6ddd2a27 | 137 | |
76e4f660 AR |
138 | if (cpu_is_offline(smp_processor_id())) |
139 | play_dead(); | |
d331e739 VP |
140 | /* |
141 | * Idle routines should keep interrupts disabled | |
142 | * from here on, until they go to idle. | |
143 | * Otherwise, idle callbacks can misfire. | |
144 | */ | |
145 | local_irq_disable(); | |
95833c83 | 146 | enter_idle(); |
81d68a96 SR |
147 | /* Don't trace irqs off for idle */ |
148 | stop_critical_timings(); | |
6ddd2a27 | 149 | pm_idle(); |
81d68a96 | 150 | start_critical_timings(); |
a15da49d AK |
151 | /* In many cases the interrupt that ended idle |
152 | has already called exit_idle. But some idle | |
153 | loops can be woken up without interrupt. */ | |
95833c83 | 154 | __exit_idle(); |
1da177e4 LT |
155 | } |
156 | ||
02290683 | 157 | tick_nohz_restart_sched_tick(); |
5bfb5d69 | 158 | preempt_enable_no_resched(); |
1da177e4 | 159 | schedule(); |
5bfb5d69 | 160 | preempt_disable(); |
1da177e4 LT |
161 | } |
162 | } | |
163 | ||
6612538c | 164 | /* Prints also some state that isn't saved in the pt_regs */ |
e2ce07c8 | 165 | void __show_regs(struct pt_regs *regs, int all) |
1da177e4 LT |
166 | { |
167 | unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L, fs, gs, shadowgs; | |
bb1995d5 | 168 | unsigned long d0, d1, d2, d3, d6, d7; |
6612538c HS |
169 | unsigned int fsindex, gsindex; |
170 | unsigned int ds, cs, es; | |
1da177e4 LT |
171 | |
172 | printk("\n"); | |
173 | print_modules(); | |
8092c654 | 174 | printk(KERN_INFO "Pid: %d, comm: %.20s %s %s %.*s\n", |
9acf23c4 | 175 | current->pid, current->comm, print_tainted(), |
96b644bd SH |
176 | init_utsname()->release, |
177 | (int)strcspn(init_utsname()->version, " "), | |
178 | init_utsname()->version); | |
8092c654 | 179 | printk(KERN_INFO "RIP: %04lx:[<%016lx>] ", regs->cs & 0xffff, regs->ip); |
aafbd7eb | 180 | printk_address(regs->ip, 1); |
8092c654 GP |
181 | printk(KERN_INFO "RSP: %04lx:%016lx EFLAGS: %08lx\n", regs->ss, |
182 | regs->sp, regs->flags); | |
183 | printk(KERN_INFO "RAX: %016lx RBX: %016lx RCX: %016lx\n", | |
65ea5b03 | 184 | regs->ax, regs->bx, regs->cx); |
8092c654 | 185 | printk(KERN_INFO "RDX: %016lx RSI: %016lx RDI: %016lx\n", |
65ea5b03 | 186 | regs->dx, regs->si, regs->di); |
8092c654 | 187 | printk(KERN_INFO "RBP: %016lx R08: %016lx R09: %016lx\n", |
65ea5b03 | 188 | regs->bp, regs->r8, regs->r9); |
8092c654 | 189 | printk(KERN_INFO "R10: %016lx R11: %016lx R12: %016lx\n", |
7de08b4e | 190 | regs->r10, regs->r11, regs->r12); |
8092c654 | 191 | printk(KERN_INFO "R13: %016lx R14: %016lx R15: %016lx\n", |
7de08b4e | 192 | regs->r13, regs->r14, regs->r15); |
1da177e4 | 193 | |
7de08b4e GP |
194 | asm("movl %%ds,%0" : "=r" (ds)); |
195 | asm("movl %%cs,%0" : "=r" (cs)); | |
196 | asm("movl %%es,%0" : "=r" (es)); | |
1da177e4 LT |
197 | asm("movl %%fs,%0" : "=r" (fsindex)); |
198 | asm("movl %%gs,%0" : "=r" (gsindex)); | |
199 | ||
200 | rdmsrl(MSR_FS_BASE, fs); | |
7de08b4e GP |
201 | rdmsrl(MSR_GS_BASE, gs); |
202 | rdmsrl(MSR_KERNEL_GS_BASE, shadowgs); | |
1da177e4 | 203 | |
e2ce07c8 PE |
204 | if (!all) |
205 | return; | |
1da177e4 | 206 | |
f51c9452 GOC |
207 | cr0 = read_cr0(); |
208 | cr2 = read_cr2(); | |
209 | cr3 = read_cr3(); | |
210 | cr4 = read_cr4(); | |
1da177e4 | 211 | |
8092c654 | 212 | printk(KERN_INFO "FS: %016lx(%04x) GS:%016lx(%04x) knlGS:%016lx\n", |
7de08b4e | 213 | fs, fsindex, gs, gsindex, shadowgs); |
8092c654 GP |
214 | printk(KERN_INFO "CS: %04x DS: %04x ES: %04x CR0: %016lx\n", cs, ds, |
215 | es, cr0); | |
216 | printk(KERN_INFO "CR2: %016lx CR3: %016lx CR4: %016lx\n", cr2, cr3, | |
217 | cr4); | |
bb1995d5 AS |
218 | |
219 | get_debugreg(d0, 0); | |
220 | get_debugreg(d1, 1); | |
221 | get_debugreg(d2, 2); | |
8092c654 | 222 | printk(KERN_INFO "DR0: %016lx DR1: %016lx DR2: %016lx\n", d0, d1, d2); |
bb1995d5 AS |
223 | get_debugreg(d3, 3); |
224 | get_debugreg(d6, 6); | |
225 | get_debugreg(d7, 7); | |
8092c654 | 226 | printk(KERN_INFO "DR3: %016lx DR6: %016lx DR7: %016lx\n", d3, d6, d7); |
1da177e4 LT |
227 | } |
228 | ||
229 | void show_regs(struct pt_regs *regs) | |
230 | { | |
8092c654 | 231 | printk(KERN_INFO "CPU %d:", smp_processor_id()); |
e2ce07c8 | 232 | __show_regs(regs, 1); |
bc850d6b | 233 | show_trace(NULL, regs, (void *)(regs + 1), regs->bp); |
1da177e4 LT |
234 | } |
235 | ||
236 | /* | |
237 | * Free current thread data structures etc.. | |
238 | */ | |
239 | void exit_thread(void) | |
240 | { | |
241 | struct task_struct *me = current; | |
242 | struct thread_struct *t = &me->thread; | |
73649dab | 243 | |
6612538c | 244 | if (me->thread.io_bitmap_ptr) { |
1da177e4 LT |
245 | struct tss_struct *tss = &per_cpu(init_tss, get_cpu()); |
246 | ||
247 | kfree(t->io_bitmap_ptr); | |
248 | t->io_bitmap_ptr = NULL; | |
d3a4f48d | 249 | clear_thread_flag(TIF_IO_BITMAP); |
1da177e4 LT |
250 | /* |
251 | * Careful, clear this in the TSS too: | |
252 | */ | |
253 | memset(tss->io_bitmap, 0xff, t->io_bitmap_max); | |
254 | t->io_bitmap_max = 0; | |
255 | put_cpu(); | |
256 | } | |
bf53de90 MM |
257 | |
258 | ds_exit_thread(current); | |
1da177e4 LT |
259 | } |
260 | ||
261 | void flush_thread(void) | |
262 | { | |
263 | struct task_struct *tsk = current; | |
1da177e4 | 264 | |
303cd153 MD |
265 | if (test_tsk_thread_flag(tsk, TIF_ABI_PENDING)) { |
266 | clear_tsk_thread_flag(tsk, TIF_ABI_PENDING); | |
267 | if (test_tsk_thread_flag(tsk, TIF_IA32)) { | |
268 | clear_tsk_thread_flag(tsk, TIF_IA32); | |
269 | } else { | |
270 | set_tsk_thread_flag(tsk, TIF_IA32); | |
4d9bc79c | 271 | current_thread_info()->status |= TS_COMPAT; |
303cd153 | 272 | } |
4d9bc79c | 273 | } |
303cd153 | 274 | clear_tsk_thread_flag(tsk, TIF_DEBUG); |
1da177e4 LT |
275 | |
276 | tsk->thread.debugreg0 = 0; | |
277 | tsk->thread.debugreg1 = 0; | |
278 | tsk->thread.debugreg2 = 0; | |
279 | tsk->thread.debugreg3 = 0; | |
280 | tsk->thread.debugreg6 = 0; | |
281 | tsk->thread.debugreg7 = 0; | |
6612538c | 282 | memset(tsk->thread.tls_array, 0, sizeof(tsk->thread.tls_array)); |
1da177e4 LT |
283 | /* |
284 | * Forget coprocessor state.. | |
285 | */ | |
75118a82 | 286 | tsk->fpu_counter = 0; |
1da177e4 LT |
287 | clear_fpu(tsk); |
288 | clear_used_math(); | |
289 | } | |
290 | ||
291 | void release_thread(struct task_struct *dead_task) | |
292 | { | |
293 | if (dead_task->mm) { | |
294 | if (dead_task->mm->context.size) { | |
295 | printk("WARNING: dead process %8s still has LDT? <%p/%d>\n", | |
296 | dead_task->comm, | |
297 | dead_task->mm->context.ldt, | |
298 | dead_task->mm->context.size); | |
299 | BUG(); | |
300 | } | |
301 | } | |
302 | } | |
303 | ||
304 | static inline void set_32bit_tls(struct task_struct *t, int tls, u32 addr) | |
305 | { | |
6612538c | 306 | struct user_desc ud = { |
1da177e4 LT |
307 | .base_addr = addr, |
308 | .limit = 0xfffff, | |
309 | .seg_32bit = 1, | |
310 | .limit_in_pages = 1, | |
311 | .useable = 1, | |
312 | }; | |
ade1af77 | 313 | struct desc_struct *desc = t->thread.tls_array; |
1da177e4 | 314 | desc += tls; |
80fbb69a | 315 | fill_ldt(desc, &ud); |
1da177e4 LT |
316 | } |
317 | ||
318 | static inline u32 read_32bit_tls(struct task_struct *t, int tls) | |
319 | { | |
91394eb0 | 320 | return get_desc_base(&t->thread.tls_array[tls]); |
1da177e4 LT |
321 | } |
322 | ||
323 | /* | |
324 | * This gets called before we allocate a new thread and copy | |
325 | * the current task into it. | |
326 | */ | |
327 | void prepare_to_copy(struct task_struct *tsk) | |
328 | { | |
329 | unlazy_fpu(tsk); | |
330 | } | |
331 | ||
65ea5b03 | 332 | int copy_thread(int nr, unsigned long clone_flags, unsigned long sp, |
1da177e4 | 333 | unsigned long unused, |
7de08b4e | 334 | struct task_struct *p, struct pt_regs *regs) |
1da177e4 LT |
335 | { |
336 | int err; | |
7de08b4e | 337 | struct pt_regs *childregs; |
1da177e4 LT |
338 | struct task_struct *me = current; |
339 | ||
a88cde13 | 340 | childregs = ((struct pt_regs *) |
57eafdc2 | 341 | (THREAD_SIZE + task_stack_page(p))) - 1; |
1da177e4 LT |
342 | *childregs = *regs; |
343 | ||
65ea5b03 PA |
344 | childregs->ax = 0; |
345 | childregs->sp = sp; | |
346 | if (sp == ~0UL) | |
347 | childregs->sp = (unsigned long)childregs; | |
1da177e4 | 348 | |
faca6227 PA |
349 | p->thread.sp = (unsigned long) childregs; |
350 | p->thread.sp0 = (unsigned long) (childregs+1); | |
351 | p->thread.usersp = me->thread.usersp; | |
1da177e4 | 352 | |
e4f17c43 | 353 | set_tsk_thread_flag(p, TIF_FORK); |
1da177e4 LT |
354 | |
355 | p->thread.fs = me->thread.fs; | |
356 | p->thread.gs = me->thread.gs; | |
357 | ||
ada85708 JF |
358 | savesegment(gs, p->thread.gsindex); |
359 | savesegment(fs, p->thread.fsindex); | |
360 | savesegment(es, p->thread.es); | |
361 | savesegment(ds, p->thread.ds); | |
1da177e4 | 362 | |
d3a4f48d | 363 | if (unlikely(test_tsk_thread_flag(me, TIF_IO_BITMAP))) { |
1da177e4 LT |
364 | p->thread.io_bitmap_ptr = kmalloc(IO_BITMAP_BYTES, GFP_KERNEL); |
365 | if (!p->thread.io_bitmap_ptr) { | |
366 | p->thread.io_bitmap_max = 0; | |
367 | return -ENOMEM; | |
368 | } | |
a88cde13 AK |
369 | memcpy(p->thread.io_bitmap_ptr, me->thread.io_bitmap_ptr, |
370 | IO_BITMAP_BYTES); | |
d3a4f48d | 371 | set_tsk_thread_flag(p, TIF_IO_BITMAP); |
6612538c | 372 | } |
1da177e4 LT |
373 | |
374 | /* | |
375 | * Set a new TLS for the child thread? | |
376 | */ | |
377 | if (clone_flags & CLONE_SETTLS) { | |
378 | #ifdef CONFIG_IA32_EMULATION | |
379 | if (test_thread_flag(TIF_IA32)) | |
efd1ca52 | 380 | err = do_set_thread_area(p, -1, |
65ea5b03 | 381 | (struct user_desc __user *)childregs->si, 0); |
7de08b4e GP |
382 | else |
383 | #endif | |
384 | err = do_arch_prctl(p, ARCH_SET_FS, childregs->r8); | |
385 | if (err) | |
1da177e4 LT |
386 | goto out; |
387 | } | |
bf53de90 MM |
388 | |
389 | ds_copy_thread(p, me); | |
390 | ||
391 | clear_tsk_thread_flag(p, TIF_DEBUGCTLMSR); | |
392 | p->thread.debugctlmsr = 0; | |
393 | ||
1da177e4 LT |
394 | err = 0; |
395 | out: | |
396 | if (err && p->thread.io_bitmap_ptr) { | |
397 | kfree(p->thread.io_bitmap_ptr); | |
398 | p->thread.io_bitmap_max = 0; | |
399 | } | |
400 | return err; | |
401 | } | |
402 | ||
513ad84b IM |
403 | void |
404 | start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp) | |
405 | { | |
ada85708 JF |
406 | loadsegment(fs, 0); |
407 | loadsegment(es, 0); | |
408 | loadsegment(ds, 0); | |
513ad84b IM |
409 | load_gs_index(0); |
410 | regs->ip = new_ip; | |
411 | regs->sp = new_sp; | |
3d1e42a7 | 412 | percpu_write(old_rsp, new_sp); |
513ad84b IM |
413 | regs->cs = __USER_CS; |
414 | regs->ss = __USER_DS; | |
415 | regs->flags = 0x200; | |
416 | set_fs(USER_DS); | |
aa283f49 SS |
417 | /* |
418 | * Free the old FP and other extended state | |
419 | */ | |
420 | free_thread_xstate(current); | |
513ad84b IM |
421 | } |
422 | EXPORT_SYMBOL_GPL(start_thread); | |
423 | ||
529e25f6 EB |
424 | static void hard_disable_TSC(void) |
425 | { | |
426 | write_cr4(read_cr4() | X86_CR4_TSD); | |
427 | } | |
428 | ||
429 | void disable_TSC(void) | |
430 | { | |
431 | preempt_disable(); | |
432 | if (!test_and_set_thread_flag(TIF_NOTSC)) | |
433 | /* | |
434 | * Must flip the CPU state synchronously with | |
435 | * TIF_NOTSC in the current running context. | |
436 | */ | |
437 | hard_disable_TSC(); | |
438 | preempt_enable(); | |
439 | } | |
440 | ||
441 | static void hard_enable_TSC(void) | |
442 | { | |
443 | write_cr4(read_cr4() & ~X86_CR4_TSD); | |
444 | } | |
445 | ||
a4928cff | 446 | static void enable_TSC(void) |
529e25f6 EB |
447 | { |
448 | preempt_disable(); | |
449 | if (test_and_clear_thread_flag(TIF_NOTSC)) | |
450 | /* | |
451 | * Must flip the CPU state synchronously with | |
452 | * TIF_NOTSC in the current running context. | |
453 | */ | |
454 | hard_enable_TSC(); | |
455 | preempt_enable(); | |
456 | } | |
457 | ||
458 | int get_tsc_mode(unsigned long adr) | |
459 | { | |
460 | unsigned int val; | |
461 | ||
462 | if (test_thread_flag(TIF_NOTSC)) | |
463 | val = PR_TSC_SIGSEGV; | |
464 | else | |
465 | val = PR_TSC_ENABLE; | |
466 | ||
467 | return put_user(val, (unsigned int __user *)adr); | |
468 | } | |
469 | ||
470 | int set_tsc_mode(unsigned int val) | |
471 | { | |
472 | if (val == PR_TSC_SIGSEGV) | |
473 | disable_TSC(); | |
474 | else if (val == PR_TSC_ENABLE) | |
475 | enable_TSC(); | |
476 | else | |
477 | return -EINVAL; | |
478 | ||
479 | return 0; | |
480 | } | |
481 | ||
1da177e4 LT |
482 | /* |
483 | * This special macro can be used to load a debugging register | |
484 | */ | |
6612538c HS |
485 | #define loaddebug(thread, r) set_debugreg(thread->debugreg ## r, r) |
486 | ||
d3a4f48d | 487 | static inline void __switch_to_xtra(struct task_struct *prev_p, |
6612538c HS |
488 | struct task_struct *next_p, |
489 | struct tss_struct *tss) | |
d3a4f48d SE |
490 | { |
491 | struct thread_struct *prev, *next; | |
492 | ||
493 | prev = &prev_p->thread, | |
494 | next = &next_p->thread; | |
495 | ||
c2724775 MM |
496 | if (test_tsk_thread_flag(next_p, TIF_DS_AREA_MSR) || |
497 | test_tsk_thread_flag(prev_p, TIF_DS_AREA_MSR)) | |
498 | ds_switch_to(prev_p, next_p); | |
499 | else if (next->debugctlmsr != prev->debugctlmsr) | |
5b0e5084 | 500 | update_debugctlmsr(next->debugctlmsr); |
7e991604 | 501 | |
d3a4f48d SE |
502 | if (test_tsk_thread_flag(next_p, TIF_DEBUG)) { |
503 | loaddebug(next, 0); | |
504 | loaddebug(next, 1); | |
505 | loaddebug(next, 2); | |
506 | loaddebug(next, 3); | |
507 | /* no 4 and 5 */ | |
508 | loaddebug(next, 6); | |
509 | loaddebug(next, 7); | |
510 | } | |
511 | ||
529e25f6 EB |
512 | if (test_tsk_thread_flag(prev_p, TIF_NOTSC) ^ |
513 | test_tsk_thread_flag(next_p, TIF_NOTSC)) { | |
514 | /* prev and next are different */ | |
515 | if (test_tsk_thread_flag(next_p, TIF_NOTSC)) | |
516 | hard_disable_TSC(); | |
517 | else | |
518 | hard_enable_TSC(); | |
519 | } | |
520 | ||
d3a4f48d SE |
521 | if (test_tsk_thread_flag(next_p, TIF_IO_BITMAP)) { |
522 | /* | |
523 | * Copy the relevant range of the IO bitmap. | |
524 | * Normally this is 128 bytes or less: | |
525 | */ | |
526 | memcpy(tss->io_bitmap, next->io_bitmap_ptr, | |
527 | max(prev->io_bitmap_max, next->io_bitmap_max)); | |
528 | } else if (test_tsk_thread_flag(prev_p, TIF_IO_BITMAP)) { | |
529 | /* | |
530 | * Clear any possible leftover bits: | |
531 | */ | |
532 | memset(tss->io_bitmap, 0xff, prev->io_bitmap_max); | |
533 | } | |
534 | } | |
535 | ||
1da177e4 LT |
536 | /* |
537 | * switch_to(x,y) should switch tasks from x to y. | |
538 | * | |
6612538c | 539 | * This could still be optimized: |
1da177e4 LT |
540 | * - fold all the options into a flag word and test it with a single test. |
541 | * - could test fs/gs bitsliced | |
099f318b AK |
542 | * |
543 | * Kprobes not supported here. Set the probe on schedule instead. | |
8b96f011 | 544 | * Function graph tracer not supported too. |
1da177e4 | 545 | */ |
8b96f011 | 546 | __notrace_funcgraph struct task_struct * |
a88cde13 | 547 | __switch_to(struct task_struct *prev_p, struct task_struct *next_p) |
1da177e4 | 548 | { |
87b935a0 JF |
549 | struct thread_struct *prev = &prev_p->thread; |
550 | struct thread_struct *next = &next_p->thread; | |
6612538c | 551 | int cpu = smp_processor_id(); |
1da177e4 | 552 | struct tss_struct *tss = &per_cpu(init_tss, cpu); |
478de5a9 | 553 | unsigned fsindex, gsindex; |
1da177e4 | 554 | |
e07e23e1 | 555 | /* we're going to use this soon, after a few expensive things */ |
7de08b4e | 556 | if (next_p->fpu_counter > 5) |
61c4628b | 557 | prefetch(next->xstate); |
e07e23e1 | 558 | |
1da177e4 LT |
559 | /* |
560 | * Reload esp0, LDT and the page table pointer: | |
561 | */ | |
7818a1e0 | 562 | load_sp0(tss, next); |
1da177e4 | 563 | |
7de08b4e | 564 | /* |
1da177e4 LT |
565 | * Switch DS and ES. |
566 | * This won't pick up thread selector changes, but I guess that is ok. | |
567 | */ | |
ada85708 | 568 | savesegment(es, prev->es); |
1da177e4 | 569 | if (unlikely(next->es | prev->es)) |
7de08b4e | 570 | loadsegment(es, next->es); |
ada85708 JF |
571 | |
572 | savesegment(ds, prev->ds); | |
1da177e4 LT |
573 | if (unlikely(next->ds | prev->ds)) |
574 | loadsegment(ds, next->ds); | |
575 | ||
478de5a9 JF |
576 | |
577 | /* We must save %fs and %gs before load_TLS() because | |
578 | * %fs and %gs may be cleared by load_TLS(). | |
579 | * | |
580 | * (e.g. xen_load_tls()) | |
581 | */ | |
582 | savesegment(fs, fsindex); | |
583 | savesegment(gs, gsindex); | |
584 | ||
1da177e4 LT |
585 | load_TLS(next, cpu); |
586 | ||
3fe0a63e JF |
587 | /* |
588 | * Leave lazy mode, flushing any hypercalls made here. | |
589 | * This must be done before restoring TLS segments so | |
590 | * the GDT and LDT are properly updated, and must be | |
591 | * done before math_state_restore, so the TS bit is up | |
592 | * to date. | |
593 | */ | |
594 | arch_leave_lazy_cpu_mode(); | |
595 | ||
7de08b4e | 596 | /* |
1da177e4 | 597 | * Switch FS and GS. |
87b935a0 JF |
598 | * |
599 | * Segment register != 0 always requires a reload. Also | |
600 | * reload when it has changed. When prev process used 64bit | |
601 | * base always reload to avoid an information leak. | |
1da177e4 | 602 | */ |
87b935a0 JF |
603 | if (unlikely(fsindex | next->fsindex | prev->fs)) { |
604 | loadsegment(fs, next->fsindex); | |
7de08b4e | 605 | /* |
87b935a0 JF |
606 | * Check if the user used a selector != 0; if yes |
607 | * clear 64bit base, since overloaded base is always | |
608 | * mapped to the Null selector | |
609 | */ | |
610 | if (fsindex) | |
7de08b4e | 611 | prev->fs = 0; |
1da177e4 | 612 | } |
87b935a0 JF |
613 | /* when next process has a 64bit base use it */ |
614 | if (next->fs) | |
615 | wrmsrl(MSR_FS_BASE, next->fs); | |
616 | prev->fsindex = fsindex; | |
617 | ||
618 | if (unlikely(gsindex | next->gsindex | prev->gs)) { | |
619 | load_gs_index(next->gsindex); | |
620 | if (gsindex) | |
7de08b4e | 621 | prev->gs = 0; |
1da177e4 | 622 | } |
87b935a0 JF |
623 | if (next->gs) |
624 | wrmsrl(MSR_KERNEL_GS_BASE, next->gs); | |
625 | prev->gsindex = gsindex; | |
1da177e4 | 626 | |
0a5ace2a AK |
627 | /* Must be after DS reload */ |
628 | unlazy_fpu(prev_p); | |
629 | ||
7de08b4e | 630 | /* |
45948d77 | 631 | * Switch the PDA and FPU contexts. |
1da177e4 | 632 | */ |
3d1e42a7 BG |
633 | prev->usersp = percpu_read(old_rsp); |
634 | percpu_write(old_rsp, next->usersp); | |
c6f5e0ac | 635 | percpu_write(current_task, next_p); |
18bd057b | 636 | |
9af45651 | 637 | percpu_write(kernel_stack, |
87b935a0 | 638 | (unsigned long)task_stack_page(next_p) + |
9af45651 | 639 | THREAD_SIZE - KERNEL_STACK_OFFSET); |
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; | |
7de08b4e | 668 | char *filename; |
1da177e4 LT |
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; | |
7de08b4e | 726 | u64 fp, ip; |
1da177e4 LT |
727 | int count = 0; |
728 | ||
7de08b4e GP |
729 | if (!p || p == current || p->state == TASK_RUNNING) |
730 | return 0; | |
57eafdc2 | 731 | stack = (unsigned long)task_stack_page(p); |
e1e23bb0 | 732 | if (p->thread.sp < stack || p->thread.sp >= stack+THREAD_SIZE) |
1da177e4 | 733 | return 0; |
faca6227 | 734 | fp = *(u64 *)(p->thread.sp); |
7de08b4e | 735 | do { |
a88cde13 | 736 | if (fp < (unsigned long)stack || |
e1e23bb0 | 737 | fp >= (unsigned long)stack+THREAD_SIZE) |
7de08b4e | 738 | return 0; |
65ea5b03 PA |
739 | ip = *(u64 *)(fp+8); |
740 | if (!in_sched_functions(ip)) | |
741 | return ip; | |
7de08b4e GP |
742 | fp = *(u64 *)fp; |
743 | } while (count++ < 16); | |
1da177e4 LT |
744 | return 0; |
745 | } | |
746 | ||
747 | long do_arch_prctl(struct task_struct *task, int code, unsigned long addr) | |
7de08b4e GP |
748 | { |
749 | int ret = 0; | |
1da177e4 LT |
750 | int doit = task == current; |
751 | int cpu; | |
752 | ||
7de08b4e | 753 | switch (code) { |
1da177e4 | 754 | case ARCH_SET_GS: |
84929801 | 755 | if (addr >= TASK_SIZE_OF(task)) |
7de08b4e | 756 | return -EPERM; |
1da177e4 | 757 | cpu = get_cpu(); |
7de08b4e | 758 | /* handle small bases via the GDT because that's faster to |
1da177e4 | 759 | switch. */ |
7de08b4e GP |
760 | if (addr <= 0xffffffff) { |
761 | set_32bit_tls(task, GS_TLS, addr); | |
762 | if (doit) { | |
1da177e4 | 763 | load_TLS(&task->thread, cpu); |
7de08b4e | 764 | load_gs_index(GS_TLS_SEL); |
1da177e4 | 765 | } |
7de08b4e | 766 | task->thread.gsindex = GS_TLS_SEL; |
1da177e4 | 767 | task->thread.gs = 0; |
7de08b4e | 768 | } else { |
1da177e4 LT |
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); | |
7de08b4e | 774 | } |
1da177e4 LT |
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; | |
7de08b4e | 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 | } |