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> |
bf53de90 | 52 | #include <asm/ds.h> |
66cb5917 | 53 | #include <asm/debugreg.h> |
1da177e4 LT |
54 | |
55 | asmlinkage extern void ret_from_fork(void); | |
56 | ||
3d1e42a7 | 57 | DEFINE_PER_CPU(unsigned long, old_rsp); |
c2558e0e | 58 | static DEFINE_PER_CPU(unsigned char, is_idle); |
3d1e42a7 | 59 | |
e041c683 | 60 | static ATOMIC_NOTIFIER_HEAD(idle_notifier); |
95833c83 AK |
61 | |
62 | void idle_notifier_register(struct notifier_block *n) | |
63 | { | |
e041c683 | 64 | atomic_notifier_chain_register(&idle_notifier, n); |
95833c83 | 65 | } |
c7d87d79 VP |
66 | EXPORT_SYMBOL_GPL(idle_notifier_register); |
67 | ||
68 | void idle_notifier_unregister(struct notifier_block *n) | |
69 | { | |
70 | atomic_notifier_chain_unregister(&idle_notifier, n); | |
71 | } | |
72 | EXPORT_SYMBOL_GPL(idle_notifier_unregister); | |
95833c83 | 73 | |
95833c83 AK |
74 | void enter_idle(void) |
75 | { | |
c2558e0e | 76 | percpu_write(is_idle, 1); |
e041c683 | 77 | atomic_notifier_call_chain(&idle_notifier, IDLE_START, NULL); |
95833c83 AK |
78 | } |
79 | ||
80 | static void __exit_idle(void) | |
81 | { | |
c2558e0e | 82 | if (x86_test_and_clear_bit_percpu(0, is_idle) == 0) |
a15da49d | 83 | return; |
e041c683 | 84 | atomic_notifier_call_chain(&idle_notifier, IDLE_END, NULL); |
95833c83 AK |
85 | } |
86 | ||
87 | /* Called from interrupts to signify idle end */ | |
88 | void exit_idle(void) | |
89 | { | |
a15da49d AK |
90 | /* idle loop has pid 0 */ |
91 | if (current->pid) | |
95833c83 AK |
92 | return; |
93 | __exit_idle(); | |
94 | } | |
95 | ||
913da64b | 96 | #ifndef CONFIG_SMP |
76e4f660 AR |
97 | static inline void play_dead(void) |
98 | { | |
99 | BUG(); | |
100 | } | |
913da64b | 101 | #endif |
76e4f660 | 102 | |
1da177e4 LT |
103 | /* |
104 | * The idle thread. There's no useful work to be | |
105 | * done, so just try to conserve power and have a | |
106 | * low exit latency (ie sit in a loop waiting for | |
107 | * somebody to say that they'd like to reschedule) | |
108 | */ | |
b10db7f0 | 109 | void cpu_idle(void) |
1da177e4 | 110 | { |
495ab9c0 | 111 | current_thread_info()->status |= TS_POLLING; |
ce22bd92 | 112 | |
ce22bd92 | 113 | /* |
5c79d2a5 TH |
114 | * If we're the non-boot CPU, nothing set the stack canary up |
115 | * for us. CPU0 already has it initialized but no harm in | |
116 | * doing it again. This is a good place for updating it, as | |
117 | * we wont ever return from this function (so the invalid | |
118 | * canaries already on the stack wont ever trigger). | |
ce22bd92 | 119 | */ |
18aa8bb1 IM |
120 | boot_init_stack_canary(); |
121 | ||
1da177e4 LT |
122 | /* endless idle loop with no priority at all */ |
123 | while (1) { | |
b8f8c3cf | 124 | tick_nohz_stop_sched_tick(1); |
1da177e4 | 125 | while (!need_resched()) { |
1da177e4 | 126 | |
1da177e4 | 127 | rmb(); |
6ddd2a27 | 128 | |
76e4f660 AR |
129 | if (cpu_is_offline(smp_processor_id())) |
130 | play_dead(); | |
d331e739 VP |
131 | /* |
132 | * Idle routines should keep interrupts disabled | |
133 | * from here on, until they go to idle. | |
134 | * Otherwise, idle callbacks can misfire. | |
135 | */ | |
136 | local_irq_disable(); | |
95833c83 | 137 | enter_idle(); |
81d68a96 SR |
138 | /* Don't trace irqs off for idle */ |
139 | stop_critical_timings(); | |
6ddd2a27 | 140 | pm_idle(); |
81d68a96 | 141 | start_critical_timings(); |
a15da49d AK |
142 | /* In many cases the interrupt that ended idle |
143 | has already called exit_idle. But some idle | |
144 | loops can be woken up without interrupt. */ | |
95833c83 | 145 | __exit_idle(); |
1da177e4 LT |
146 | } |
147 | ||
02290683 | 148 | tick_nohz_restart_sched_tick(); |
5bfb5d69 | 149 | preempt_enable_no_resched(); |
1da177e4 | 150 | schedule(); |
5bfb5d69 | 151 | preempt_disable(); |
1da177e4 LT |
152 | } |
153 | } | |
154 | ||
6612538c | 155 | /* Prints also some state that isn't saved in the pt_regs */ |
e2ce07c8 | 156 | void __show_regs(struct pt_regs *regs, int all) |
1da177e4 LT |
157 | { |
158 | unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L, fs, gs, shadowgs; | |
bb1995d5 | 159 | unsigned long d0, d1, d2, d3, d6, d7; |
6612538c HS |
160 | unsigned int fsindex, gsindex; |
161 | unsigned int ds, cs, es; | |
814e2c84 AI |
162 | |
163 | show_regs_common(); | |
d015a092 | 164 | printk(KERN_DEFAULT "RIP: %04lx:[<%016lx>] ", regs->cs & 0xffff, regs->ip); |
aafbd7eb | 165 | printk_address(regs->ip, 1); |
d015a092 | 166 | printk(KERN_DEFAULT "RSP: %04lx:%016lx EFLAGS: %08lx\n", regs->ss, |
8092c654 | 167 | regs->sp, regs->flags); |
d015a092 | 168 | printk(KERN_DEFAULT "RAX: %016lx RBX: %016lx RCX: %016lx\n", |
65ea5b03 | 169 | regs->ax, regs->bx, regs->cx); |
d015a092 | 170 | printk(KERN_DEFAULT "RDX: %016lx RSI: %016lx RDI: %016lx\n", |
65ea5b03 | 171 | regs->dx, regs->si, regs->di); |
d015a092 | 172 | printk(KERN_DEFAULT "RBP: %016lx R08: %016lx R09: %016lx\n", |
65ea5b03 | 173 | regs->bp, regs->r8, regs->r9); |
d015a092 | 174 | printk(KERN_DEFAULT "R10: %016lx R11: %016lx R12: %016lx\n", |
7de08b4e | 175 | regs->r10, regs->r11, regs->r12); |
d015a092 | 176 | printk(KERN_DEFAULT "R13: %016lx R14: %016lx R15: %016lx\n", |
7de08b4e | 177 | regs->r13, regs->r14, regs->r15); |
1da177e4 | 178 | |
7de08b4e GP |
179 | asm("movl %%ds,%0" : "=r" (ds)); |
180 | asm("movl %%cs,%0" : "=r" (cs)); | |
181 | asm("movl %%es,%0" : "=r" (es)); | |
1da177e4 LT |
182 | asm("movl %%fs,%0" : "=r" (fsindex)); |
183 | asm("movl %%gs,%0" : "=r" (gsindex)); | |
184 | ||
185 | rdmsrl(MSR_FS_BASE, fs); | |
7de08b4e GP |
186 | rdmsrl(MSR_GS_BASE, gs); |
187 | rdmsrl(MSR_KERNEL_GS_BASE, shadowgs); | |
1da177e4 | 188 | |
e2ce07c8 PE |
189 | if (!all) |
190 | return; | |
1da177e4 | 191 | |
f51c9452 GOC |
192 | cr0 = read_cr0(); |
193 | cr2 = read_cr2(); | |
194 | cr3 = read_cr3(); | |
195 | cr4 = read_cr4(); | |
1da177e4 | 196 | |
d015a092 | 197 | printk(KERN_DEFAULT "FS: %016lx(%04x) GS:%016lx(%04x) knlGS:%016lx\n", |
7de08b4e | 198 | fs, fsindex, gs, gsindex, shadowgs); |
d015a092 | 199 | printk(KERN_DEFAULT "CS: %04x DS: %04x ES: %04x CR0: %016lx\n", cs, ds, |
8092c654 | 200 | es, cr0); |
d015a092 | 201 | printk(KERN_DEFAULT "CR2: %016lx CR3: %016lx CR4: %016lx\n", cr2, cr3, |
8092c654 | 202 | cr4); |
bb1995d5 AS |
203 | |
204 | get_debugreg(d0, 0); | |
205 | get_debugreg(d1, 1); | |
206 | get_debugreg(d2, 2); | |
d015a092 | 207 | printk(KERN_DEFAULT "DR0: %016lx DR1: %016lx DR2: %016lx\n", d0, d1, d2); |
bb1995d5 AS |
208 | get_debugreg(d3, 3); |
209 | get_debugreg(d6, 6); | |
210 | get_debugreg(d7, 7); | |
d015a092 | 211 | printk(KERN_DEFAULT "DR3: %016lx DR6: %016lx DR7: %016lx\n", d3, d6, d7); |
1da177e4 LT |
212 | } |
213 | ||
214 | void show_regs(struct pt_regs *regs) | |
215 | { | |
a489ca35 | 216 | show_registers(regs); |
bc850d6b | 217 | show_trace(NULL, regs, (void *)(regs + 1), regs->bp); |
1da177e4 LT |
218 | } |
219 | ||
1da177e4 LT |
220 | void release_thread(struct task_struct *dead_task) |
221 | { | |
222 | if (dead_task->mm) { | |
223 | if (dead_task->mm->context.size) { | |
224 | printk("WARNING: dead process %8s still has LDT? <%p/%d>\n", | |
225 | dead_task->comm, | |
226 | dead_task->mm->context.ldt, | |
227 | dead_task->mm->context.size); | |
228 | BUG(); | |
229 | } | |
230 | } | |
231 | } | |
232 | ||
233 | static inline void set_32bit_tls(struct task_struct *t, int tls, u32 addr) | |
234 | { | |
6612538c | 235 | struct user_desc ud = { |
1da177e4 LT |
236 | .base_addr = addr, |
237 | .limit = 0xfffff, | |
238 | .seg_32bit = 1, | |
239 | .limit_in_pages = 1, | |
240 | .useable = 1, | |
241 | }; | |
ade1af77 | 242 | struct desc_struct *desc = t->thread.tls_array; |
1da177e4 | 243 | desc += tls; |
80fbb69a | 244 | fill_ldt(desc, &ud); |
1da177e4 LT |
245 | } |
246 | ||
247 | static inline u32 read_32bit_tls(struct task_struct *t, int tls) | |
248 | { | |
91394eb0 | 249 | return get_desc_base(&t->thread.tls_array[tls]); |
1da177e4 LT |
250 | } |
251 | ||
252 | /* | |
253 | * This gets called before we allocate a new thread and copy | |
254 | * the current task into it. | |
255 | */ | |
256 | void prepare_to_copy(struct task_struct *tsk) | |
257 | { | |
258 | unlazy_fpu(tsk); | |
259 | } | |
260 | ||
6f2c55b8 | 261 | int copy_thread(unsigned long clone_flags, unsigned long sp, |
1da177e4 | 262 | unsigned long unused, |
7de08b4e | 263 | struct task_struct *p, struct pt_regs *regs) |
1da177e4 LT |
264 | { |
265 | int err; | |
7de08b4e | 266 | struct pt_regs *childregs; |
1da177e4 LT |
267 | struct task_struct *me = current; |
268 | ||
a88cde13 | 269 | childregs = ((struct pt_regs *) |
57eafdc2 | 270 | (THREAD_SIZE + task_stack_page(p))) - 1; |
1da177e4 LT |
271 | *childregs = *regs; |
272 | ||
65ea5b03 | 273 | childregs->ax = 0; |
fa4b8f84 BG |
274 | if (user_mode(regs)) |
275 | childregs->sp = sp; | |
276 | else | |
65ea5b03 | 277 | childregs->sp = (unsigned long)childregs; |
1da177e4 | 278 | |
faca6227 PA |
279 | p->thread.sp = (unsigned long) childregs; |
280 | p->thread.sp0 = (unsigned long) (childregs+1); | |
281 | p->thread.usersp = me->thread.usersp; | |
1da177e4 | 282 | |
e4f17c43 | 283 | set_tsk_thread_flag(p, TIF_FORK); |
1da177e4 LT |
284 | |
285 | p->thread.fs = me->thread.fs; | |
286 | p->thread.gs = me->thread.gs; | |
66cb5917 | 287 | p->thread.io_bitmap_ptr = NULL; |
1da177e4 | 288 | |
ada85708 JF |
289 | savesegment(gs, p->thread.gsindex); |
290 | savesegment(fs, p->thread.fsindex); | |
291 | savesegment(es, p->thread.es); | |
292 | savesegment(ds, p->thread.ds); | |
1da177e4 | 293 | |
66cb5917 | 294 | err = -ENOMEM; |
24f1e32c | 295 | memset(p->thread.ptrace_bps, 0, sizeof(p->thread.ptrace_bps)); |
66cb5917 | 296 | |
d3a4f48d | 297 | if (unlikely(test_tsk_thread_flag(me, TIF_IO_BITMAP))) { |
1da177e4 LT |
298 | p->thread.io_bitmap_ptr = kmalloc(IO_BITMAP_BYTES, GFP_KERNEL); |
299 | if (!p->thread.io_bitmap_ptr) { | |
300 | p->thread.io_bitmap_max = 0; | |
301 | return -ENOMEM; | |
302 | } | |
a88cde13 AK |
303 | memcpy(p->thread.io_bitmap_ptr, me->thread.io_bitmap_ptr, |
304 | IO_BITMAP_BYTES); | |
d3a4f48d | 305 | set_tsk_thread_flag(p, TIF_IO_BITMAP); |
6612538c | 306 | } |
1da177e4 LT |
307 | |
308 | /* | |
309 | * Set a new TLS for the child thread? | |
310 | */ | |
311 | if (clone_flags & CLONE_SETTLS) { | |
312 | #ifdef CONFIG_IA32_EMULATION | |
313 | if (test_thread_flag(TIF_IA32)) | |
efd1ca52 | 314 | err = do_set_thread_area(p, -1, |
65ea5b03 | 315 | (struct user_desc __user *)childregs->si, 0); |
7de08b4e GP |
316 | else |
317 | #endif | |
318 | err = do_arch_prctl(p, ARCH_SET_FS, childregs->r8); | |
319 | if (err) | |
1da177e4 LT |
320 | goto out; |
321 | } | |
bf53de90 | 322 | |
2311f0de MM |
323 | clear_tsk_thread_flag(p, TIF_DS_AREA_MSR); |
324 | p->thread.ds_ctx = NULL; | |
bf53de90 MM |
325 | |
326 | clear_tsk_thread_flag(p, TIF_DEBUGCTLMSR); | |
327 | p->thread.debugctlmsr = 0; | |
328 | ||
1da177e4 LT |
329 | err = 0; |
330 | out: | |
331 | if (err && p->thread.io_bitmap_ptr) { | |
332 | kfree(p->thread.io_bitmap_ptr); | |
333 | p->thread.io_bitmap_max = 0; | |
334 | } | |
66cb5917 | 335 | |
1da177e4 LT |
336 | return err; |
337 | } | |
338 | ||
e634d8fc PA |
339 | static void |
340 | start_thread_common(struct pt_regs *regs, unsigned long new_ip, | |
341 | unsigned long new_sp, | |
342 | unsigned int _cs, unsigned int _ss, unsigned int _ds) | |
513ad84b | 343 | { |
ada85708 | 344 | loadsegment(fs, 0); |
e634d8fc PA |
345 | loadsegment(es, _ds); |
346 | loadsegment(ds, _ds); | |
513ad84b IM |
347 | load_gs_index(0); |
348 | regs->ip = new_ip; | |
349 | regs->sp = new_sp; | |
3d1e42a7 | 350 | percpu_write(old_rsp, new_sp); |
e634d8fc PA |
351 | regs->cs = _cs; |
352 | regs->ss = _ss; | |
a6f05a6a | 353 | regs->flags = X86_EFLAGS_IF; |
513ad84b | 354 | set_fs(USER_DS); |
aa283f49 SS |
355 | /* |
356 | * Free the old FP and other extended state | |
357 | */ | |
358 | free_thread_xstate(current); | |
513ad84b | 359 | } |
e634d8fc PA |
360 | |
361 | void | |
362 | start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp) | |
363 | { | |
364 | start_thread_common(regs, new_ip, new_sp, | |
365 | __USER_CS, __USER_DS, 0); | |
366 | } | |
513ad84b | 367 | |
a6f05a6a PA |
368 | #ifdef CONFIG_IA32_EMULATION |
369 | void start_thread_ia32(struct pt_regs *regs, u32 new_ip, u32 new_sp) | |
370 | { | |
e634d8fc PA |
371 | start_thread_common(regs, new_ip, new_sp, |
372 | __USER32_CS, __USER32_DS, __USER32_DS); | |
a6f05a6a PA |
373 | } |
374 | #endif | |
513ad84b | 375 | |
1da177e4 LT |
376 | /* |
377 | * switch_to(x,y) should switch tasks from x to y. | |
378 | * | |
6612538c | 379 | * This could still be optimized: |
1da177e4 LT |
380 | * - fold all the options into a flag word and test it with a single test. |
381 | * - could test fs/gs bitsliced | |
099f318b AK |
382 | * |
383 | * Kprobes not supported here. Set the probe on schedule instead. | |
8b96f011 | 384 | * Function graph tracer not supported too. |
1da177e4 | 385 | */ |
8b96f011 | 386 | __notrace_funcgraph struct task_struct * |
a88cde13 | 387 | __switch_to(struct task_struct *prev_p, struct task_struct *next_p) |
1da177e4 | 388 | { |
87b935a0 JF |
389 | struct thread_struct *prev = &prev_p->thread; |
390 | struct thread_struct *next = &next_p->thread; | |
6612538c | 391 | int cpu = smp_processor_id(); |
1da177e4 | 392 | struct tss_struct *tss = &per_cpu(init_tss, cpu); |
478de5a9 | 393 | unsigned fsindex, gsindex; |
17950c5b JF |
394 | bool preload_fpu; |
395 | ||
396 | /* | |
397 | * If the task has used fpu the last 5 timeslices, just do a full | |
398 | * restore of the math state immediately to avoid the trap; the | |
399 | * chances of needing FPU soon are obviously high now | |
400 | */ | |
401 | preload_fpu = tsk_used_math(next_p) && next_p->fpu_counter > 5; | |
1da177e4 | 402 | |
e07e23e1 | 403 | /* we're going to use this soon, after a few expensive things */ |
17950c5b | 404 | if (preload_fpu) |
61c4628b | 405 | prefetch(next->xstate); |
e07e23e1 | 406 | |
1da177e4 LT |
407 | /* |
408 | * Reload esp0, LDT and the page table pointer: | |
409 | */ | |
7818a1e0 | 410 | load_sp0(tss, next); |
1da177e4 | 411 | |
7de08b4e | 412 | /* |
1da177e4 LT |
413 | * Switch DS and ES. |
414 | * This won't pick up thread selector changes, but I guess that is ok. | |
415 | */ | |
ada85708 | 416 | savesegment(es, prev->es); |
1da177e4 | 417 | if (unlikely(next->es | prev->es)) |
7de08b4e | 418 | loadsegment(es, next->es); |
ada85708 JF |
419 | |
420 | savesegment(ds, prev->ds); | |
1da177e4 LT |
421 | if (unlikely(next->ds | prev->ds)) |
422 | loadsegment(ds, next->ds); | |
423 | ||
478de5a9 JF |
424 | |
425 | /* We must save %fs and %gs before load_TLS() because | |
426 | * %fs and %gs may be cleared by load_TLS(). | |
427 | * | |
428 | * (e.g. xen_load_tls()) | |
429 | */ | |
430 | savesegment(fs, fsindex); | |
431 | savesegment(gs, gsindex); | |
432 | ||
1da177e4 LT |
433 | load_TLS(next, cpu); |
434 | ||
16d9dbf0 JF |
435 | /* Must be after DS reload */ |
436 | unlazy_fpu(prev_p); | |
437 | ||
17950c5b JF |
438 | /* Make sure cpu is ready for new context */ |
439 | if (preload_fpu) | |
440 | clts(); | |
441 | ||
3fe0a63e JF |
442 | /* |
443 | * Leave lazy mode, flushing any hypercalls made here. | |
444 | * This must be done before restoring TLS segments so | |
445 | * the GDT and LDT are properly updated, and must be | |
446 | * done before math_state_restore, so the TS bit is up | |
447 | * to date. | |
448 | */ | |
224101ed | 449 | arch_end_context_switch(next_p); |
3fe0a63e | 450 | |
7de08b4e | 451 | /* |
1da177e4 | 452 | * Switch FS and GS. |
87b935a0 JF |
453 | * |
454 | * Segment register != 0 always requires a reload. Also | |
455 | * reload when it has changed. When prev process used 64bit | |
456 | * base always reload to avoid an information leak. | |
1da177e4 | 457 | */ |
87b935a0 JF |
458 | if (unlikely(fsindex | next->fsindex | prev->fs)) { |
459 | loadsegment(fs, next->fsindex); | |
7de08b4e | 460 | /* |
87b935a0 JF |
461 | * Check if the user used a selector != 0; if yes |
462 | * clear 64bit base, since overloaded base is always | |
463 | * mapped to the Null selector | |
464 | */ | |
465 | if (fsindex) | |
7de08b4e | 466 | prev->fs = 0; |
1da177e4 | 467 | } |
87b935a0 JF |
468 | /* when next process has a 64bit base use it */ |
469 | if (next->fs) | |
470 | wrmsrl(MSR_FS_BASE, next->fs); | |
471 | prev->fsindex = fsindex; | |
472 | ||
473 | if (unlikely(gsindex | next->gsindex | prev->gs)) { | |
474 | load_gs_index(next->gsindex); | |
475 | if (gsindex) | |
7de08b4e | 476 | prev->gs = 0; |
1da177e4 | 477 | } |
87b935a0 JF |
478 | if (next->gs) |
479 | wrmsrl(MSR_KERNEL_GS_BASE, next->gs); | |
480 | prev->gsindex = gsindex; | |
1da177e4 | 481 | |
7de08b4e | 482 | /* |
45948d77 | 483 | * Switch the PDA and FPU contexts. |
1da177e4 | 484 | */ |
3d1e42a7 BG |
485 | prev->usersp = percpu_read(old_rsp); |
486 | percpu_write(old_rsp, next->usersp); | |
c6f5e0ac | 487 | percpu_write(current_task, next_p); |
18bd057b | 488 | |
9af45651 | 489 | percpu_write(kernel_stack, |
87b935a0 | 490 | (unsigned long)task_stack_page(next_p) + |
9af45651 | 491 | THREAD_SIZE - KERNEL_STACK_OFFSET); |
1da177e4 LT |
492 | |
493 | /* | |
d3a4f48d | 494 | * Now maybe reload the debug registers and handle I/O bitmaps |
1da177e4 | 495 | */ |
eee3af4a MM |
496 | if (unlikely(task_thread_info(next_p)->flags & _TIF_WORK_CTXSW_NEXT || |
497 | task_thread_info(prev_p)->flags & _TIF_WORK_CTXSW_PREV)) | |
d3a4f48d | 498 | __switch_to_xtra(prev_p, next_p, tss); |
1da177e4 | 499 | |
17950c5b JF |
500 | /* |
501 | * Preload the FPU context, now that we've determined that the | |
502 | * task is likely to be using it. | |
e07e23e1 | 503 | */ |
17950c5b JF |
504 | if (preload_fpu) |
505 | __math_state_restore(); | |
66cb5917 | 506 | |
1da177e4 LT |
507 | return prev_p; |
508 | } | |
509 | ||
1da177e4 LT |
510 | void set_personality_64bit(void) |
511 | { | |
512 | /* inherit personality from parent */ | |
513 | ||
514 | /* Make sure to be in 64bit mode */ | |
6612538c | 515 | clear_thread_flag(TIF_IA32); |
1da177e4 LT |
516 | |
517 | /* TBD: overwrites user setup. Should have two bits. | |
518 | But 64bit processes have always behaved this way, | |
519 | so it's not too bad. The main problem is just that | |
6612538c | 520 | 32bit childs are affected again. */ |
1da177e4 LT |
521 | current->personality &= ~READ_IMPLIES_EXEC; |
522 | } | |
523 | ||
05d43ed8 PA |
524 | void set_personality_ia32(void) |
525 | { | |
526 | /* inherit personality from parent */ | |
527 | ||
528 | /* Make sure to be in 32bit mode */ | |
529 | set_thread_flag(TIF_IA32); | |
1252f238 | 530 | current->personality |= force_personality32; |
05d43ed8 PA |
531 | |
532 | /* Prepare the first "return" to user space */ | |
533 | current_thread_info()->status |= TS_COMPAT; | |
534 | } | |
535 | ||
1da177e4 LT |
536 | unsigned long get_wchan(struct task_struct *p) |
537 | { | |
538 | unsigned long stack; | |
7de08b4e | 539 | u64 fp, ip; |
1da177e4 LT |
540 | int count = 0; |
541 | ||
7de08b4e GP |
542 | if (!p || p == current || p->state == TASK_RUNNING) |
543 | return 0; | |
57eafdc2 | 544 | stack = (unsigned long)task_stack_page(p); |
e1e23bb0 | 545 | if (p->thread.sp < stack || p->thread.sp >= stack+THREAD_SIZE) |
1da177e4 | 546 | return 0; |
faca6227 | 547 | fp = *(u64 *)(p->thread.sp); |
7de08b4e | 548 | do { |
a88cde13 | 549 | if (fp < (unsigned long)stack || |
e1e23bb0 | 550 | fp >= (unsigned long)stack+THREAD_SIZE) |
7de08b4e | 551 | return 0; |
65ea5b03 PA |
552 | ip = *(u64 *)(fp+8); |
553 | if (!in_sched_functions(ip)) | |
554 | return ip; | |
7de08b4e GP |
555 | fp = *(u64 *)fp; |
556 | } while (count++ < 16); | |
1da177e4 LT |
557 | return 0; |
558 | } | |
559 | ||
560 | long do_arch_prctl(struct task_struct *task, int code, unsigned long addr) | |
7de08b4e GP |
561 | { |
562 | int ret = 0; | |
1da177e4 LT |
563 | int doit = task == current; |
564 | int cpu; | |
565 | ||
7de08b4e | 566 | switch (code) { |
1da177e4 | 567 | case ARCH_SET_GS: |
84929801 | 568 | if (addr >= TASK_SIZE_OF(task)) |
7de08b4e | 569 | return -EPERM; |
1da177e4 | 570 | cpu = get_cpu(); |
7de08b4e | 571 | /* handle small bases via the GDT because that's faster to |
1da177e4 | 572 | switch. */ |
7de08b4e GP |
573 | if (addr <= 0xffffffff) { |
574 | set_32bit_tls(task, GS_TLS, addr); | |
575 | if (doit) { | |
1da177e4 | 576 | load_TLS(&task->thread, cpu); |
7de08b4e | 577 | load_gs_index(GS_TLS_SEL); |
1da177e4 | 578 | } |
7de08b4e | 579 | task->thread.gsindex = GS_TLS_SEL; |
1da177e4 | 580 | task->thread.gs = 0; |
7de08b4e | 581 | } else { |
1da177e4 LT |
582 | task->thread.gsindex = 0; |
583 | task->thread.gs = addr; | |
584 | if (doit) { | |
a88cde13 AK |
585 | load_gs_index(0); |
586 | ret = checking_wrmsrl(MSR_KERNEL_GS_BASE, addr); | |
7de08b4e | 587 | } |
1da177e4 LT |
588 | } |
589 | put_cpu(); | |
590 | break; | |
591 | case ARCH_SET_FS: | |
592 | /* Not strictly needed for fs, but do it for symmetry | |
593 | with gs */ | |
84929801 | 594 | if (addr >= TASK_SIZE_OF(task)) |
6612538c | 595 | return -EPERM; |
1da177e4 | 596 | cpu = get_cpu(); |
6612538c | 597 | /* handle small bases via the GDT because that's faster to |
1da177e4 | 598 | switch. */ |
6612538c | 599 | if (addr <= 0xffffffff) { |
1da177e4 | 600 | set_32bit_tls(task, FS_TLS, addr); |
6612538c HS |
601 | if (doit) { |
602 | load_TLS(&task->thread, cpu); | |
ada85708 | 603 | loadsegment(fs, FS_TLS_SEL); |
1da177e4 LT |
604 | } |
605 | task->thread.fsindex = FS_TLS_SEL; | |
606 | task->thread.fs = 0; | |
6612538c | 607 | } else { |
1da177e4 LT |
608 | task->thread.fsindex = 0; |
609 | task->thread.fs = addr; | |
610 | if (doit) { | |
611 | /* set the selector to 0 to not confuse | |
612 | __switch_to */ | |
ada85708 | 613 | loadsegment(fs, 0); |
a88cde13 | 614 | ret = checking_wrmsrl(MSR_FS_BASE, addr); |
1da177e4 LT |
615 | } |
616 | } | |
617 | put_cpu(); | |
618 | break; | |
6612538c HS |
619 | case ARCH_GET_FS: { |
620 | unsigned long base; | |
1da177e4 LT |
621 | if (task->thread.fsindex == FS_TLS_SEL) |
622 | base = read_32bit_tls(task, FS_TLS); | |
a88cde13 | 623 | else if (doit) |
1da177e4 | 624 | rdmsrl(MSR_FS_BASE, base); |
a88cde13 | 625 | else |
1da177e4 | 626 | base = task->thread.fs; |
6612538c HS |
627 | ret = put_user(base, (unsigned long __user *)addr); |
628 | break; | |
1da177e4 | 629 | } |
6612538c | 630 | case ARCH_GET_GS: { |
1da177e4 | 631 | unsigned long base; |
97c2803c | 632 | unsigned gsindex; |
1da177e4 LT |
633 | if (task->thread.gsindex == GS_TLS_SEL) |
634 | base = read_32bit_tls(task, GS_TLS); | |
97c2803c | 635 | else if (doit) { |
ada85708 | 636 | savesegment(gs, gsindex); |
97c2803c JB |
637 | if (gsindex) |
638 | rdmsrl(MSR_KERNEL_GS_BASE, base); | |
639 | else | |
640 | base = task->thread.gs; | |
7de08b4e | 641 | } else |
1da177e4 | 642 | base = task->thread.gs; |
6612538c | 643 | ret = put_user(base, (unsigned long __user *)addr); |
1da177e4 LT |
644 | break; |
645 | } | |
646 | ||
647 | default: | |
648 | ret = -EINVAL; | |
649 | break; | |
6612538c | 650 | } |
1da177e4 | 651 | |
6612538c HS |
652 | return ret; |
653 | } | |
1da177e4 LT |
654 | |
655 | long sys_arch_prctl(int code, unsigned long addr) | |
656 | { | |
657 | return do_arch_prctl(current, code, addr); | |
1da177e4 LT |
658 | } |
659 | ||
89240ba0 SS |
660 | unsigned long KSTK_ESP(struct task_struct *task) |
661 | { | |
662 | return (test_tsk_thread_flag(task, TIF_IA32)) ? | |
663 | (task_pt_regs(task)->sp) : ((task)->thread.usersp); | |
664 | } |