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 | ||
42059429 | 17 | #include <linux/stackprotector.h> |
76e4f660 | 18 | #include <linux/cpu.h> |
1da177e4 LT |
19 | #include <linux/errno.h> |
20 | #include <linux/sched.h> | |
6612538c | 21 | #include <linux/fs.h> |
1da177e4 LT |
22 | #include <linux/kernel.h> |
23 | #include <linux/mm.h> | |
24 | #include <linux/elfcore.h> | |
25 | #include <linux/smp.h> | |
26 | #include <linux/slab.h> | |
27 | #include <linux/user.h> | |
1da177e4 LT |
28 | #include <linux/interrupt.h> |
29 | #include <linux/delay.h> | |
6612538c | 30 | #include <linux/module.h> |
1da177e4 | 31 | #include <linux/ptrace.h> |
95833c83 | 32 | #include <linux/notifier.h> |
c6fd91f0 | 33 | #include <linux/kprobes.h> |
1eeb66a1 | 34 | #include <linux/kdebug.h> |
02290683 | 35 | #include <linux/tick.h> |
529e25f6 | 36 | #include <linux/prctl.h> |
7de08b4e GP |
37 | #include <linux/uaccess.h> |
38 | #include <linux/io.h> | |
8b96f011 | 39 | #include <linux/ftrace.h> |
1da177e4 | 40 | |
1da177e4 LT |
41 | #include <asm/pgtable.h> |
42 | #include <asm/system.h> | |
1da177e4 LT |
43 | #include <asm/processor.h> |
44 | #include <asm/i387.h> | |
45 | #include <asm/mmu_context.h> | |
1da177e4 | 46 | #include <asm/prctl.h> |
1da177e4 LT |
47 | #include <asm/desc.h> |
48 | #include <asm/proto.h> | |
49 | #include <asm/ia32.h> | |
95833c83 | 50 | #include <asm/idle.h> |
bbc1f698 | 51 | #include <asm/syscalls.h> |
66cb5917 | 52 | #include <asm/debugreg.h> |
1da177e4 LT |
53 | |
54 | asmlinkage extern void ret_from_fork(void); | |
55 | ||
3d1e42a7 | 56 | DEFINE_PER_CPU(unsigned long, old_rsp); |
c2558e0e | 57 | static DEFINE_PER_CPU(unsigned char, is_idle); |
3d1e42a7 | 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 | } |
c7d87d79 VP |
65 | EXPORT_SYMBOL_GPL(idle_notifier_register); |
66 | ||
67 | void idle_notifier_unregister(struct notifier_block *n) | |
68 | { | |
69 | atomic_notifier_chain_unregister(&idle_notifier, n); | |
70 | } | |
71 | EXPORT_SYMBOL_GPL(idle_notifier_unregister); | |
95833c83 | 72 | |
95833c83 AK |
73 | void enter_idle(void) |
74 | { | |
c2558e0e | 75 | percpu_write(is_idle, 1); |
e041c683 | 76 | atomic_notifier_call_chain(&idle_notifier, IDLE_START, NULL); |
95833c83 AK |
77 | } |
78 | ||
79 | static void __exit_idle(void) | |
80 | { | |
c2558e0e | 81 | if (x86_test_and_clear_bit_percpu(0, is_idle) == 0) |
a15da49d | 82 | return; |
e041c683 | 83 | atomic_notifier_call_chain(&idle_notifier, IDLE_END, NULL); |
95833c83 AK |
84 | } |
85 | ||
86 | /* Called from interrupts to signify idle end */ | |
87 | void exit_idle(void) | |
88 | { | |
a15da49d AK |
89 | /* idle loop has pid 0 */ |
90 | if (current->pid) | |
95833c83 AK |
91 | return; |
92 | __exit_idle(); | |
93 | } | |
94 | ||
913da64b | 95 | #ifndef CONFIG_SMP |
76e4f660 AR |
96 | static inline void play_dead(void) |
97 | { | |
98 | BUG(); | |
99 | } | |
913da64b | 100 | #endif |
76e4f660 | 101 | |
1da177e4 LT |
102 | /* |
103 | * The idle thread. There's no useful work to be | |
104 | * done, so just try to conserve power and have a | |
105 | * low exit latency (ie sit in a loop waiting for | |
106 | * somebody to say that they'd like to reschedule) | |
107 | */ | |
b10db7f0 | 108 | void cpu_idle(void) |
1da177e4 | 109 | { |
495ab9c0 | 110 | current_thread_info()->status |= TS_POLLING; |
ce22bd92 | 111 | |
ce22bd92 | 112 | /* |
5c79d2a5 TH |
113 | * If we're the non-boot CPU, nothing set the stack canary up |
114 | * for us. CPU0 already has it initialized but no harm in | |
115 | * doing it again. This is a good place for updating it, as | |
116 | * we wont ever return from this function (so the invalid | |
117 | * canaries already on the stack wont ever trigger). | |
ce22bd92 | 118 | */ |
18aa8bb1 IM |
119 | boot_init_stack_canary(); |
120 | ||
1da177e4 LT |
121 | /* endless idle loop with no priority at all */ |
122 | while (1) { | |
b8f8c3cf | 123 | tick_nohz_stop_sched_tick(1); |
1da177e4 | 124 | while (!need_resched()) { |
1da177e4 | 125 | |
1da177e4 | 126 | rmb(); |
6ddd2a27 | 127 | |
76e4f660 AR |
128 | if (cpu_is_offline(smp_processor_id())) |
129 | play_dead(); | |
d331e739 VP |
130 | /* |
131 | * Idle routines should keep interrupts disabled | |
132 | * from here on, until they go to idle. | |
133 | * Otherwise, idle callbacks can misfire. | |
134 | */ | |
135 | local_irq_disable(); | |
95833c83 | 136 | enter_idle(); |
81d68a96 SR |
137 | /* Don't trace irqs off for idle */ |
138 | stop_critical_timings(); | |
6ddd2a27 | 139 | pm_idle(); |
81d68a96 | 140 | start_critical_timings(); |
a15da49d AK |
141 | /* In many cases the interrupt that ended idle |
142 | has already called exit_idle. But some idle | |
143 | loops can be woken up without interrupt. */ | |
95833c83 | 144 | __exit_idle(); |
1da177e4 LT |
145 | } |
146 | ||
02290683 | 147 | tick_nohz_restart_sched_tick(); |
5bfb5d69 | 148 | preempt_enable_no_resched(); |
1da177e4 | 149 | schedule(); |
5bfb5d69 | 150 | preempt_disable(); |
1da177e4 LT |
151 | } |
152 | } | |
153 | ||
6612538c | 154 | /* Prints also some state that isn't saved in the pt_regs */ |
e2ce07c8 | 155 | void __show_regs(struct pt_regs *regs, int all) |
1da177e4 LT |
156 | { |
157 | unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L, fs, gs, shadowgs; | |
bb1995d5 | 158 | unsigned long d0, d1, d2, d3, d6, d7; |
6612538c HS |
159 | unsigned int fsindex, gsindex; |
160 | unsigned int ds, cs, es; | |
814e2c84 AI |
161 | |
162 | show_regs_common(); | |
d015a092 | 163 | printk(KERN_DEFAULT "RIP: %04lx:[<%016lx>] ", regs->cs & 0xffff, regs->ip); |
aafbd7eb | 164 | printk_address(regs->ip, 1); |
d015a092 | 165 | printk(KERN_DEFAULT "RSP: %04lx:%016lx EFLAGS: %08lx\n", regs->ss, |
8092c654 | 166 | regs->sp, regs->flags); |
d015a092 | 167 | printk(KERN_DEFAULT "RAX: %016lx RBX: %016lx RCX: %016lx\n", |
65ea5b03 | 168 | regs->ax, regs->bx, regs->cx); |
d015a092 | 169 | printk(KERN_DEFAULT "RDX: %016lx RSI: %016lx RDI: %016lx\n", |
65ea5b03 | 170 | regs->dx, regs->si, regs->di); |
d015a092 | 171 | printk(KERN_DEFAULT "RBP: %016lx R08: %016lx R09: %016lx\n", |
65ea5b03 | 172 | regs->bp, regs->r8, regs->r9); |
d015a092 | 173 | printk(KERN_DEFAULT "R10: %016lx R11: %016lx R12: %016lx\n", |
7de08b4e | 174 | regs->r10, regs->r11, regs->r12); |
d015a092 | 175 | printk(KERN_DEFAULT "R13: %016lx R14: %016lx R15: %016lx\n", |
7de08b4e | 176 | regs->r13, regs->r14, regs->r15); |
1da177e4 | 177 | |
7de08b4e GP |
178 | asm("movl %%ds,%0" : "=r" (ds)); |
179 | asm("movl %%cs,%0" : "=r" (cs)); | |
180 | asm("movl %%es,%0" : "=r" (es)); | |
1da177e4 LT |
181 | asm("movl %%fs,%0" : "=r" (fsindex)); |
182 | asm("movl %%gs,%0" : "=r" (gsindex)); | |
183 | ||
184 | rdmsrl(MSR_FS_BASE, fs); | |
7de08b4e GP |
185 | rdmsrl(MSR_GS_BASE, gs); |
186 | rdmsrl(MSR_KERNEL_GS_BASE, shadowgs); | |
1da177e4 | 187 | |
e2ce07c8 PE |
188 | if (!all) |
189 | return; | |
1da177e4 | 190 | |
f51c9452 GOC |
191 | cr0 = read_cr0(); |
192 | cr2 = read_cr2(); | |
193 | cr3 = read_cr3(); | |
194 | cr4 = read_cr4(); | |
1da177e4 | 195 | |
d015a092 | 196 | printk(KERN_DEFAULT "FS: %016lx(%04x) GS:%016lx(%04x) knlGS:%016lx\n", |
7de08b4e | 197 | fs, fsindex, gs, gsindex, shadowgs); |
d015a092 | 198 | printk(KERN_DEFAULT "CS: %04x DS: %04x ES: %04x CR0: %016lx\n", cs, ds, |
8092c654 | 199 | es, cr0); |
d015a092 | 200 | printk(KERN_DEFAULT "CR2: %016lx CR3: %016lx CR4: %016lx\n", cr2, cr3, |
8092c654 | 201 | cr4); |
bb1995d5 AS |
202 | |
203 | get_debugreg(d0, 0); | |
204 | get_debugreg(d1, 1); | |
205 | get_debugreg(d2, 2); | |
d015a092 | 206 | printk(KERN_DEFAULT "DR0: %016lx DR1: %016lx DR2: %016lx\n", d0, d1, d2); |
bb1995d5 AS |
207 | get_debugreg(d3, 3); |
208 | get_debugreg(d6, 6); | |
209 | get_debugreg(d7, 7); | |
d015a092 | 210 | printk(KERN_DEFAULT "DR3: %016lx DR6: %016lx DR7: %016lx\n", d3, d6, d7); |
1da177e4 LT |
211 | } |
212 | ||
1da177e4 LT |
213 | void release_thread(struct task_struct *dead_task) |
214 | { | |
215 | if (dead_task->mm) { | |
216 | if (dead_task->mm->context.size) { | |
217 | printk("WARNING: dead process %8s still has LDT? <%p/%d>\n", | |
218 | dead_task->comm, | |
219 | dead_task->mm->context.ldt, | |
220 | dead_task->mm->context.size); | |
221 | BUG(); | |
222 | } | |
223 | } | |
224 | } | |
225 | ||
226 | static inline void set_32bit_tls(struct task_struct *t, int tls, u32 addr) | |
227 | { | |
6612538c | 228 | struct user_desc ud = { |
1da177e4 LT |
229 | .base_addr = addr, |
230 | .limit = 0xfffff, | |
231 | .seg_32bit = 1, | |
232 | .limit_in_pages = 1, | |
233 | .useable = 1, | |
234 | }; | |
ade1af77 | 235 | struct desc_struct *desc = t->thread.tls_array; |
1da177e4 | 236 | desc += tls; |
80fbb69a | 237 | fill_ldt(desc, &ud); |
1da177e4 LT |
238 | } |
239 | ||
240 | static inline u32 read_32bit_tls(struct task_struct *t, int tls) | |
241 | { | |
91394eb0 | 242 | return get_desc_base(&t->thread.tls_array[tls]); |
1da177e4 LT |
243 | } |
244 | ||
245 | /* | |
246 | * This gets called before we allocate a new thread and copy | |
247 | * the current task into it. | |
248 | */ | |
249 | void prepare_to_copy(struct task_struct *tsk) | |
250 | { | |
251 | unlazy_fpu(tsk); | |
252 | } | |
253 | ||
6f2c55b8 | 254 | int copy_thread(unsigned long clone_flags, unsigned long sp, |
1da177e4 | 255 | unsigned long unused, |
7de08b4e | 256 | struct task_struct *p, struct pt_regs *regs) |
1da177e4 LT |
257 | { |
258 | int err; | |
7de08b4e | 259 | struct pt_regs *childregs; |
1da177e4 LT |
260 | struct task_struct *me = current; |
261 | ||
a88cde13 | 262 | childregs = ((struct pt_regs *) |
57eafdc2 | 263 | (THREAD_SIZE + task_stack_page(p))) - 1; |
1da177e4 LT |
264 | *childregs = *regs; |
265 | ||
65ea5b03 | 266 | childregs->ax = 0; |
fa4b8f84 BG |
267 | if (user_mode(regs)) |
268 | childregs->sp = sp; | |
269 | else | |
65ea5b03 | 270 | childregs->sp = (unsigned long)childregs; |
1da177e4 | 271 | |
faca6227 PA |
272 | p->thread.sp = (unsigned long) childregs; |
273 | p->thread.sp0 = (unsigned long) (childregs+1); | |
274 | p->thread.usersp = me->thread.usersp; | |
1da177e4 | 275 | |
e4f17c43 | 276 | set_tsk_thread_flag(p, TIF_FORK); |
1da177e4 | 277 | |
66cb5917 | 278 | p->thread.io_bitmap_ptr = NULL; |
1da177e4 | 279 | |
ada85708 | 280 | savesegment(gs, p->thread.gsindex); |
7ce5a2b9 | 281 | p->thread.gs = p->thread.gsindex ? 0 : me->thread.gs; |
ada85708 | 282 | savesegment(fs, p->thread.fsindex); |
7ce5a2b9 | 283 | p->thread.fs = p->thread.fsindex ? 0 : me->thread.fs; |
ada85708 JF |
284 | savesegment(es, p->thread.es); |
285 | savesegment(ds, p->thread.ds); | |
1da177e4 | 286 | |
66cb5917 | 287 | err = -ENOMEM; |
24f1e32c | 288 | memset(p->thread.ptrace_bps, 0, sizeof(p->thread.ptrace_bps)); |
66cb5917 | 289 | |
d3a4f48d | 290 | if (unlikely(test_tsk_thread_flag(me, TIF_IO_BITMAP))) { |
1da177e4 LT |
291 | p->thread.io_bitmap_ptr = kmalloc(IO_BITMAP_BYTES, GFP_KERNEL); |
292 | if (!p->thread.io_bitmap_ptr) { | |
293 | p->thread.io_bitmap_max = 0; | |
294 | return -ENOMEM; | |
295 | } | |
a88cde13 AK |
296 | memcpy(p->thread.io_bitmap_ptr, me->thread.io_bitmap_ptr, |
297 | IO_BITMAP_BYTES); | |
d3a4f48d | 298 | set_tsk_thread_flag(p, TIF_IO_BITMAP); |
6612538c | 299 | } |
1da177e4 LT |
300 | |
301 | /* | |
302 | * Set a new TLS for the child thread? | |
303 | */ | |
304 | if (clone_flags & CLONE_SETTLS) { | |
305 | #ifdef CONFIG_IA32_EMULATION | |
306 | if (test_thread_flag(TIF_IA32)) | |
efd1ca52 | 307 | err = do_set_thread_area(p, -1, |
65ea5b03 | 308 | (struct user_desc __user *)childregs->si, 0); |
7de08b4e GP |
309 | else |
310 | #endif | |
311 | err = do_arch_prctl(p, ARCH_SET_FS, childregs->r8); | |
312 | if (err) | |
1da177e4 LT |
313 | goto out; |
314 | } | |
315 | err = 0; | |
316 | out: | |
317 | if (err && p->thread.io_bitmap_ptr) { | |
318 | kfree(p->thread.io_bitmap_ptr); | |
319 | p->thread.io_bitmap_max = 0; | |
320 | } | |
66cb5917 | 321 | |
1da177e4 LT |
322 | return err; |
323 | } | |
324 | ||
e634d8fc PA |
325 | static void |
326 | start_thread_common(struct pt_regs *regs, unsigned long new_ip, | |
327 | unsigned long new_sp, | |
328 | unsigned int _cs, unsigned int _ss, unsigned int _ds) | |
513ad84b | 329 | { |
ada85708 | 330 | loadsegment(fs, 0); |
e634d8fc PA |
331 | loadsegment(es, _ds); |
332 | loadsegment(ds, _ds); | |
513ad84b IM |
333 | load_gs_index(0); |
334 | regs->ip = new_ip; | |
335 | regs->sp = new_sp; | |
3d1e42a7 | 336 | percpu_write(old_rsp, new_sp); |
e634d8fc PA |
337 | regs->cs = _cs; |
338 | regs->ss = _ss; | |
a6f05a6a | 339 | regs->flags = X86_EFLAGS_IF; |
513ad84b | 340 | set_fs(USER_DS); |
aa283f49 SS |
341 | /* |
342 | * Free the old FP and other extended state | |
343 | */ | |
344 | free_thread_xstate(current); | |
513ad84b | 345 | } |
e634d8fc PA |
346 | |
347 | void | |
348 | start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp) | |
349 | { | |
350 | start_thread_common(regs, new_ip, new_sp, | |
351 | __USER_CS, __USER_DS, 0); | |
352 | } | |
513ad84b | 353 | |
a6f05a6a PA |
354 | #ifdef CONFIG_IA32_EMULATION |
355 | void start_thread_ia32(struct pt_regs *regs, u32 new_ip, u32 new_sp) | |
356 | { | |
e634d8fc PA |
357 | start_thread_common(regs, new_ip, new_sp, |
358 | __USER32_CS, __USER32_DS, __USER32_DS); | |
a6f05a6a PA |
359 | } |
360 | #endif | |
513ad84b | 361 | |
1da177e4 LT |
362 | /* |
363 | * switch_to(x,y) should switch tasks from x to y. | |
364 | * | |
6612538c | 365 | * This could still be optimized: |
1da177e4 LT |
366 | * - fold all the options into a flag word and test it with a single test. |
367 | * - could test fs/gs bitsliced | |
099f318b AK |
368 | * |
369 | * Kprobes not supported here. Set the probe on schedule instead. | |
8b96f011 | 370 | * Function graph tracer not supported too. |
1da177e4 | 371 | */ |
8b96f011 | 372 | __notrace_funcgraph struct task_struct * |
a88cde13 | 373 | __switch_to(struct task_struct *prev_p, struct task_struct *next_p) |
1da177e4 | 374 | { |
87b935a0 JF |
375 | struct thread_struct *prev = &prev_p->thread; |
376 | struct thread_struct *next = &next_p->thread; | |
6612538c | 377 | int cpu = smp_processor_id(); |
1da177e4 | 378 | struct tss_struct *tss = &per_cpu(init_tss, cpu); |
478de5a9 | 379 | unsigned fsindex, gsindex; |
17950c5b JF |
380 | bool preload_fpu; |
381 | ||
382 | /* | |
383 | * If the task has used fpu the last 5 timeslices, just do a full | |
384 | * restore of the math state immediately to avoid the trap; the | |
385 | * chances of needing FPU soon are obviously high now | |
386 | */ | |
387 | preload_fpu = tsk_used_math(next_p) && next_p->fpu_counter > 5; | |
1da177e4 | 388 | |
e07e23e1 | 389 | /* we're going to use this soon, after a few expensive things */ |
17950c5b | 390 | if (preload_fpu) |
86603283 | 391 | prefetch(next->fpu.state); |
e07e23e1 | 392 | |
1da177e4 LT |
393 | /* |
394 | * Reload esp0, LDT and the page table pointer: | |
395 | */ | |
7818a1e0 | 396 | load_sp0(tss, next); |
1da177e4 | 397 | |
7de08b4e | 398 | /* |
1da177e4 LT |
399 | * Switch DS and ES. |
400 | * This won't pick up thread selector changes, but I guess that is ok. | |
401 | */ | |
ada85708 | 402 | savesegment(es, prev->es); |
1da177e4 | 403 | if (unlikely(next->es | prev->es)) |
7de08b4e | 404 | loadsegment(es, next->es); |
ada85708 JF |
405 | |
406 | savesegment(ds, prev->ds); | |
1da177e4 LT |
407 | if (unlikely(next->ds | prev->ds)) |
408 | loadsegment(ds, next->ds); | |
409 | ||
478de5a9 JF |
410 | |
411 | /* We must save %fs and %gs before load_TLS() because | |
412 | * %fs and %gs may be cleared by load_TLS(). | |
413 | * | |
414 | * (e.g. xen_load_tls()) | |
415 | */ | |
416 | savesegment(fs, fsindex); | |
417 | savesegment(gs, gsindex); | |
418 | ||
1da177e4 LT |
419 | load_TLS(next, cpu); |
420 | ||
16d9dbf0 JF |
421 | /* Must be after DS reload */ |
422 | unlazy_fpu(prev_p); | |
423 | ||
17950c5b JF |
424 | /* Make sure cpu is ready for new context */ |
425 | if (preload_fpu) | |
426 | clts(); | |
427 | ||
3fe0a63e JF |
428 | /* |
429 | * Leave lazy mode, flushing any hypercalls made here. | |
430 | * This must be done before restoring TLS segments so | |
431 | * the GDT and LDT are properly updated, and must be | |
432 | * done before math_state_restore, so the TS bit is up | |
433 | * to date. | |
434 | */ | |
224101ed | 435 | arch_end_context_switch(next_p); |
3fe0a63e | 436 | |
7de08b4e | 437 | /* |
1da177e4 | 438 | * Switch FS and GS. |
87b935a0 JF |
439 | * |
440 | * Segment register != 0 always requires a reload. Also | |
441 | * reload when it has changed. When prev process used 64bit | |
442 | * base always reload to avoid an information leak. | |
1da177e4 | 443 | */ |
87b935a0 JF |
444 | if (unlikely(fsindex | next->fsindex | prev->fs)) { |
445 | loadsegment(fs, next->fsindex); | |
7de08b4e | 446 | /* |
87b935a0 JF |
447 | * Check if the user used a selector != 0; if yes |
448 | * clear 64bit base, since overloaded base is always | |
449 | * mapped to the Null selector | |
450 | */ | |
451 | if (fsindex) | |
7de08b4e | 452 | prev->fs = 0; |
1da177e4 | 453 | } |
87b935a0 JF |
454 | /* when next process has a 64bit base use it */ |
455 | if (next->fs) | |
456 | wrmsrl(MSR_FS_BASE, next->fs); | |
457 | prev->fsindex = fsindex; | |
458 | ||
459 | if (unlikely(gsindex | next->gsindex | prev->gs)) { | |
460 | load_gs_index(next->gsindex); | |
461 | if (gsindex) | |
7de08b4e | 462 | prev->gs = 0; |
1da177e4 | 463 | } |
87b935a0 JF |
464 | if (next->gs) |
465 | wrmsrl(MSR_KERNEL_GS_BASE, next->gs); | |
466 | prev->gsindex = gsindex; | |
1da177e4 | 467 | |
7de08b4e | 468 | /* |
45948d77 | 469 | * Switch the PDA and FPU contexts. |
1da177e4 | 470 | */ |
3d1e42a7 BG |
471 | prev->usersp = percpu_read(old_rsp); |
472 | percpu_write(old_rsp, next->usersp); | |
c6f5e0ac | 473 | percpu_write(current_task, next_p); |
18bd057b | 474 | |
9af45651 | 475 | percpu_write(kernel_stack, |
87b935a0 | 476 | (unsigned long)task_stack_page(next_p) + |
9af45651 | 477 | THREAD_SIZE - KERNEL_STACK_OFFSET); |
1da177e4 LT |
478 | |
479 | /* | |
d3a4f48d | 480 | * Now maybe reload the debug registers and handle I/O bitmaps |
1da177e4 | 481 | */ |
eee3af4a MM |
482 | if (unlikely(task_thread_info(next_p)->flags & _TIF_WORK_CTXSW_NEXT || |
483 | task_thread_info(prev_p)->flags & _TIF_WORK_CTXSW_PREV)) | |
d3a4f48d | 484 | __switch_to_xtra(prev_p, next_p, tss); |
1da177e4 | 485 | |
17950c5b JF |
486 | /* |
487 | * Preload the FPU context, now that we've determined that the | |
488 | * task is likely to be using it. | |
e07e23e1 | 489 | */ |
17950c5b JF |
490 | if (preload_fpu) |
491 | __math_state_restore(); | |
66cb5917 | 492 | |
1da177e4 LT |
493 | return prev_p; |
494 | } | |
495 | ||
1da177e4 LT |
496 | void set_personality_64bit(void) |
497 | { | |
498 | /* inherit personality from parent */ | |
499 | ||
500 | /* Make sure to be in 64bit mode */ | |
6612538c | 501 | clear_thread_flag(TIF_IA32); |
1da177e4 LT |
502 | |
503 | /* TBD: overwrites user setup. Should have two bits. | |
504 | But 64bit processes have always behaved this way, | |
505 | so it's not too bad. The main problem is just that | |
6612538c | 506 | 32bit childs are affected again. */ |
1da177e4 LT |
507 | current->personality &= ~READ_IMPLIES_EXEC; |
508 | } | |
509 | ||
05d43ed8 PA |
510 | void set_personality_ia32(void) |
511 | { | |
512 | /* inherit personality from parent */ | |
513 | ||
514 | /* Make sure to be in 32bit mode */ | |
515 | set_thread_flag(TIF_IA32); | |
1252f238 | 516 | current->personality |= force_personality32; |
05d43ed8 PA |
517 | |
518 | /* Prepare the first "return" to user space */ | |
519 | current_thread_info()->status |= TS_COMPAT; | |
520 | } | |
521 | ||
1da177e4 LT |
522 | unsigned long get_wchan(struct task_struct *p) |
523 | { | |
524 | unsigned long stack; | |
7de08b4e | 525 | u64 fp, ip; |
1da177e4 LT |
526 | int count = 0; |
527 | ||
7de08b4e GP |
528 | if (!p || p == current || p->state == TASK_RUNNING) |
529 | return 0; | |
57eafdc2 | 530 | stack = (unsigned long)task_stack_page(p); |
e1e23bb0 | 531 | if (p->thread.sp < stack || p->thread.sp >= stack+THREAD_SIZE) |
1da177e4 | 532 | return 0; |
faca6227 | 533 | fp = *(u64 *)(p->thread.sp); |
7de08b4e | 534 | do { |
a88cde13 | 535 | if (fp < (unsigned long)stack || |
e1e23bb0 | 536 | fp >= (unsigned long)stack+THREAD_SIZE) |
7de08b4e | 537 | return 0; |
65ea5b03 PA |
538 | ip = *(u64 *)(fp+8); |
539 | if (!in_sched_functions(ip)) | |
540 | return ip; | |
7de08b4e GP |
541 | fp = *(u64 *)fp; |
542 | } while (count++ < 16); | |
1da177e4 LT |
543 | return 0; |
544 | } | |
545 | ||
546 | long do_arch_prctl(struct task_struct *task, int code, unsigned long addr) | |
7de08b4e GP |
547 | { |
548 | int ret = 0; | |
1da177e4 LT |
549 | int doit = task == current; |
550 | int cpu; | |
551 | ||
7de08b4e | 552 | switch (code) { |
1da177e4 | 553 | case ARCH_SET_GS: |
84929801 | 554 | if (addr >= TASK_SIZE_OF(task)) |
7de08b4e | 555 | return -EPERM; |
1da177e4 | 556 | cpu = get_cpu(); |
7de08b4e | 557 | /* handle small bases via the GDT because that's faster to |
1da177e4 | 558 | switch. */ |
7de08b4e GP |
559 | if (addr <= 0xffffffff) { |
560 | set_32bit_tls(task, GS_TLS, addr); | |
561 | if (doit) { | |
1da177e4 | 562 | load_TLS(&task->thread, cpu); |
7de08b4e | 563 | load_gs_index(GS_TLS_SEL); |
1da177e4 | 564 | } |
7de08b4e | 565 | task->thread.gsindex = GS_TLS_SEL; |
1da177e4 | 566 | task->thread.gs = 0; |
7de08b4e | 567 | } else { |
1da177e4 LT |
568 | task->thread.gsindex = 0; |
569 | task->thread.gs = addr; | |
570 | if (doit) { | |
a88cde13 AK |
571 | load_gs_index(0); |
572 | ret = checking_wrmsrl(MSR_KERNEL_GS_BASE, addr); | |
7de08b4e | 573 | } |
1da177e4 LT |
574 | } |
575 | put_cpu(); | |
576 | break; | |
577 | case ARCH_SET_FS: | |
578 | /* Not strictly needed for fs, but do it for symmetry | |
579 | with gs */ | |
84929801 | 580 | if (addr >= TASK_SIZE_OF(task)) |
6612538c | 581 | return -EPERM; |
1da177e4 | 582 | cpu = get_cpu(); |
6612538c | 583 | /* handle small bases via the GDT because that's faster to |
1da177e4 | 584 | switch. */ |
6612538c | 585 | if (addr <= 0xffffffff) { |
1da177e4 | 586 | set_32bit_tls(task, FS_TLS, addr); |
6612538c HS |
587 | if (doit) { |
588 | load_TLS(&task->thread, cpu); | |
ada85708 | 589 | loadsegment(fs, FS_TLS_SEL); |
1da177e4 LT |
590 | } |
591 | task->thread.fsindex = FS_TLS_SEL; | |
592 | task->thread.fs = 0; | |
6612538c | 593 | } else { |
1da177e4 LT |
594 | task->thread.fsindex = 0; |
595 | task->thread.fs = addr; | |
596 | if (doit) { | |
597 | /* set the selector to 0 to not confuse | |
598 | __switch_to */ | |
ada85708 | 599 | loadsegment(fs, 0); |
a88cde13 | 600 | ret = checking_wrmsrl(MSR_FS_BASE, addr); |
1da177e4 LT |
601 | } |
602 | } | |
603 | put_cpu(); | |
604 | break; | |
6612538c HS |
605 | case ARCH_GET_FS: { |
606 | unsigned long base; | |
1da177e4 LT |
607 | if (task->thread.fsindex == FS_TLS_SEL) |
608 | base = read_32bit_tls(task, FS_TLS); | |
a88cde13 | 609 | else if (doit) |
1da177e4 | 610 | rdmsrl(MSR_FS_BASE, base); |
a88cde13 | 611 | else |
1da177e4 | 612 | base = task->thread.fs; |
6612538c HS |
613 | ret = put_user(base, (unsigned long __user *)addr); |
614 | break; | |
1da177e4 | 615 | } |
6612538c | 616 | case ARCH_GET_GS: { |
1da177e4 | 617 | unsigned long base; |
97c2803c | 618 | unsigned gsindex; |
1da177e4 LT |
619 | if (task->thread.gsindex == GS_TLS_SEL) |
620 | base = read_32bit_tls(task, GS_TLS); | |
97c2803c | 621 | else if (doit) { |
ada85708 | 622 | savesegment(gs, gsindex); |
97c2803c JB |
623 | if (gsindex) |
624 | rdmsrl(MSR_KERNEL_GS_BASE, base); | |
625 | else | |
626 | base = task->thread.gs; | |
7de08b4e | 627 | } else |
1da177e4 | 628 | base = task->thread.gs; |
6612538c | 629 | ret = put_user(base, (unsigned long __user *)addr); |
1da177e4 LT |
630 | break; |
631 | } | |
632 | ||
633 | default: | |
634 | ret = -EINVAL; | |
635 | break; | |
6612538c | 636 | } |
1da177e4 | 637 | |
6612538c HS |
638 | return ret; |
639 | } | |
1da177e4 LT |
640 | |
641 | long sys_arch_prctl(int code, unsigned long addr) | |
642 | { | |
643 | return do_arch_prctl(current, code, addr); | |
1da177e4 LT |
644 | } |
645 | ||
89240ba0 SS |
646 | unsigned long KSTK_ESP(struct task_struct *task) |
647 | { | |
648 | return (test_tsk_thread_flag(task, TIF_IA32)) ? | |
649 | (task_pt_regs(task)->sp) : ((task)->thread.usersp); | |
650 | } |