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> |
48ec4d95 | 44 | #include <linux/dmi.h> |
1da177e4 | 45 | |
1da177e4 LT |
46 | #include <asm/pgtable.h> |
47 | #include <asm/system.h> | |
1da177e4 LT |
48 | #include <asm/processor.h> |
49 | #include <asm/i387.h> | |
50 | #include <asm/mmu_context.h> | |
1da177e4 | 51 | #include <asm/prctl.h> |
1da177e4 LT |
52 | #include <asm/desc.h> |
53 | #include <asm/proto.h> | |
54 | #include <asm/ia32.h> | |
95833c83 | 55 | #include <asm/idle.h> |
bbc1f698 | 56 | #include <asm/syscalls.h> |
bf53de90 | 57 | #include <asm/ds.h> |
66cb5917 P |
58 | #include <asm/debugreg.h> |
59 | #include <asm/hw_breakpoint.h> | |
1da177e4 LT |
60 | |
61 | asmlinkage extern void ret_from_fork(void); | |
62 | ||
c6f5e0ac BG |
63 | DEFINE_PER_CPU(struct task_struct *, current_task) = &init_task; |
64 | EXPORT_PER_CPU_SYMBOL(current_task); | |
65 | ||
3d1e42a7 | 66 | DEFINE_PER_CPU(unsigned long, old_rsp); |
c2558e0e | 67 | static DEFINE_PER_CPU(unsigned char, is_idle); |
3d1e42a7 | 68 | |
1da177e4 LT |
69 | unsigned long kernel_thread_flags = CLONE_VM | CLONE_UNTRACED; |
70 | ||
e041c683 | 71 | static ATOMIC_NOTIFIER_HEAD(idle_notifier); |
95833c83 AK |
72 | |
73 | void idle_notifier_register(struct notifier_block *n) | |
74 | { | |
e041c683 | 75 | atomic_notifier_chain_register(&idle_notifier, n); |
95833c83 | 76 | } |
c7d87d79 VP |
77 | EXPORT_SYMBOL_GPL(idle_notifier_register); |
78 | ||
79 | void idle_notifier_unregister(struct notifier_block *n) | |
80 | { | |
81 | atomic_notifier_chain_unregister(&idle_notifier, n); | |
82 | } | |
83 | EXPORT_SYMBOL_GPL(idle_notifier_unregister); | |
95833c83 | 84 | |
95833c83 AK |
85 | void enter_idle(void) |
86 | { | |
c2558e0e | 87 | percpu_write(is_idle, 1); |
e041c683 | 88 | atomic_notifier_call_chain(&idle_notifier, IDLE_START, NULL); |
95833c83 AK |
89 | } |
90 | ||
91 | static void __exit_idle(void) | |
92 | { | |
c2558e0e | 93 | if (x86_test_and_clear_bit_percpu(0, is_idle) == 0) |
a15da49d | 94 | return; |
e041c683 | 95 | atomic_notifier_call_chain(&idle_notifier, IDLE_END, NULL); |
95833c83 AK |
96 | } |
97 | ||
98 | /* Called from interrupts to signify idle end */ | |
99 | void exit_idle(void) | |
100 | { | |
a15da49d AK |
101 | /* idle loop has pid 0 */ |
102 | if (current->pid) | |
95833c83 AK |
103 | return; |
104 | __exit_idle(); | |
105 | } | |
106 | ||
913da64b | 107 | #ifndef CONFIG_SMP |
76e4f660 AR |
108 | static inline void play_dead(void) |
109 | { | |
110 | BUG(); | |
111 | } | |
913da64b | 112 | #endif |
76e4f660 | 113 | |
1da177e4 LT |
114 | /* |
115 | * The idle thread. There's no useful work to be | |
116 | * done, so just try to conserve power and have a | |
117 | * low exit latency (ie sit in a loop waiting for | |
118 | * somebody to say that they'd like to reschedule) | |
119 | */ | |
b10db7f0 | 120 | void cpu_idle(void) |
1da177e4 | 121 | { |
495ab9c0 | 122 | current_thread_info()->status |= TS_POLLING; |
ce22bd92 | 123 | |
ce22bd92 | 124 | /* |
5c79d2a5 TH |
125 | * If we're the non-boot CPU, nothing set the stack canary up |
126 | * for us. CPU0 already has it initialized but no harm in | |
127 | * doing it again. This is a good place for updating it, as | |
128 | * we wont ever return from this function (so the invalid | |
129 | * canaries already on the stack wont ever trigger). | |
ce22bd92 | 130 | */ |
18aa8bb1 IM |
131 | boot_init_stack_canary(); |
132 | ||
1da177e4 LT |
133 | /* endless idle loop with no priority at all */ |
134 | while (1) { | |
b8f8c3cf | 135 | tick_nohz_stop_sched_tick(1); |
1da177e4 | 136 | while (!need_resched()) { |
1da177e4 | 137 | |
1da177e4 | 138 | rmb(); |
6ddd2a27 | 139 | |
76e4f660 AR |
140 | if (cpu_is_offline(smp_processor_id())) |
141 | play_dead(); | |
d331e739 VP |
142 | /* |
143 | * Idle routines should keep interrupts disabled | |
144 | * from here on, until they go to idle. | |
145 | * Otherwise, idle callbacks can misfire. | |
146 | */ | |
147 | local_irq_disable(); | |
95833c83 | 148 | enter_idle(); |
81d68a96 SR |
149 | /* Don't trace irqs off for idle */ |
150 | stop_critical_timings(); | |
6ddd2a27 | 151 | pm_idle(); |
81d68a96 | 152 | start_critical_timings(); |
a15da49d AK |
153 | /* In many cases the interrupt that ended idle |
154 | has already called exit_idle. But some idle | |
155 | loops can be woken up without interrupt. */ | |
95833c83 | 156 | __exit_idle(); |
1da177e4 LT |
157 | } |
158 | ||
02290683 | 159 | tick_nohz_restart_sched_tick(); |
5bfb5d69 | 160 | preempt_enable_no_resched(); |
1da177e4 | 161 | schedule(); |
5bfb5d69 | 162 | preempt_disable(); |
1da177e4 LT |
163 | } |
164 | } | |
165 | ||
6612538c | 166 | /* Prints also some state that isn't saved in the pt_regs */ |
e2ce07c8 | 167 | void __show_regs(struct pt_regs *regs, int all) |
1da177e4 LT |
168 | { |
169 | unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L, fs, gs, shadowgs; | |
bb1995d5 | 170 | unsigned long d0, d1, d2, d3, d6, d7; |
6612538c HS |
171 | unsigned int fsindex, gsindex; |
172 | unsigned int ds, cs, es; | |
48ec4d95 | 173 | const char *board; |
1da177e4 LT |
174 | |
175 | printk("\n"); | |
176 | print_modules(); | |
48ec4d95 KM |
177 | board = dmi_get_system_info(DMI_PRODUCT_NAME); |
178 | if (!board) | |
179 | board = ""; | |
180 | printk(KERN_INFO "Pid: %d, comm: %.20s %s %s %.*s %s\n", | |
9acf23c4 | 181 | current->pid, current->comm, print_tainted(), |
96b644bd SH |
182 | init_utsname()->release, |
183 | (int)strcspn(init_utsname()->version, " "), | |
48ec4d95 | 184 | init_utsname()->version, board); |
8092c654 | 185 | printk(KERN_INFO "RIP: %04lx:[<%016lx>] ", regs->cs & 0xffff, regs->ip); |
aafbd7eb | 186 | printk_address(regs->ip, 1); |
8092c654 GP |
187 | printk(KERN_INFO "RSP: %04lx:%016lx EFLAGS: %08lx\n", regs->ss, |
188 | regs->sp, regs->flags); | |
189 | printk(KERN_INFO "RAX: %016lx RBX: %016lx RCX: %016lx\n", | |
65ea5b03 | 190 | regs->ax, regs->bx, regs->cx); |
8092c654 | 191 | printk(KERN_INFO "RDX: %016lx RSI: %016lx RDI: %016lx\n", |
65ea5b03 | 192 | regs->dx, regs->si, regs->di); |
8092c654 | 193 | printk(KERN_INFO "RBP: %016lx R08: %016lx R09: %016lx\n", |
65ea5b03 | 194 | regs->bp, regs->r8, regs->r9); |
8092c654 | 195 | printk(KERN_INFO "R10: %016lx R11: %016lx R12: %016lx\n", |
7de08b4e | 196 | regs->r10, regs->r11, regs->r12); |
8092c654 | 197 | printk(KERN_INFO "R13: %016lx R14: %016lx R15: %016lx\n", |
7de08b4e | 198 | regs->r13, regs->r14, regs->r15); |
1da177e4 | 199 | |
7de08b4e GP |
200 | asm("movl %%ds,%0" : "=r" (ds)); |
201 | asm("movl %%cs,%0" : "=r" (cs)); | |
202 | asm("movl %%es,%0" : "=r" (es)); | |
1da177e4 LT |
203 | asm("movl %%fs,%0" : "=r" (fsindex)); |
204 | asm("movl %%gs,%0" : "=r" (gsindex)); | |
205 | ||
206 | rdmsrl(MSR_FS_BASE, fs); | |
7de08b4e GP |
207 | rdmsrl(MSR_GS_BASE, gs); |
208 | rdmsrl(MSR_KERNEL_GS_BASE, shadowgs); | |
1da177e4 | 209 | |
e2ce07c8 PE |
210 | if (!all) |
211 | return; | |
1da177e4 | 212 | |
f51c9452 GOC |
213 | cr0 = read_cr0(); |
214 | cr2 = read_cr2(); | |
215 | cr3 = read_cr3(); | |
216 | cr4 = read_cr4(); | |
1da177e4 | 217 | |
8092c654 | 218 | printk(KERN_INFO "FS: %016lx(%04x) GS:%016lx(%04x) knlGS:%016lx\n", |
7de08b4e | 219 | fs, fsindex, gs, gsindex, shadowgs); |
8092c654 GP |
220 | printk(KERN_INFO "CS: %04x DS: %04x ES: %04x CR0: %016lx\n", cs, ds, |
221 | es, cr0); | |
222 | printk(KERN_INFO "CR2: %016lx CR3: %016lx CR4: %016lx\n", cr2, cr3, | |
223 | cr4); | |
bb1995d5 AS |
224 | |
225 | get_debugreg(d0, 0); | |
226 | get_debugreg(d1, 1); | |
227 | get_debugreg(d2, 2); | |
8092c654 | 228 | printk(KERN_INFO "DR0: %016lx DR1: %016lx DR2: %016lx\n", d0, d1, d2); |
bb1995d5 AS |
229 | get_debugreg(d3, 3); |
230 | get_debugreg(d6, 6); | |
231 | get_debugreg(d7, 7); | |
8092c654 | 232 | printk(KERN_INFO "DR3: %016lx DR6: %016lx DR7: %016lx\n", d3, d6, d7); |
1da177e4 LT |
233 | } |
234 | ||
235 | void show_regs(struct pt_regs *regs) | |
236 | { | |
8092c654 | 237 | printk(KERN_INFO "CPU %d:", smp_processor_id()); |
e2ce07c8 | 238 | __show_regs(regs, 1); |
bc850d6b | 239 | show_trace(NULL, regs, (void *)(regs + 1), regs->bp); |
1da177e4 LT |
240 | } |
241 | ||
1da177e4 LT |
242 | void release_thread(struct task_struct *dead_task) |
243 | { | |
244 | if (dead_task->mm) { | |
245 | if (dead_task->mm->context.size) { | |
246 | printk("WARNING: dead process %8s still has LDT? <%p/%d>\n", | |
247 | dead_task->comm, | |
248 | dead_task->mm->context.ldt, | |
249 | dead_task->mm->context.size); | |
250 | BUG(); | |
251 | } | |
252 | } | |
66cb5917 P |
253 | if (unlikely(dead_task->thread.debugreg7)) |
254 | flush_thread_hw_breakpoint(dead_task); | |
1da177e4 LT |
255 | } |
256 | ||
257 | static inline void set_32bit_tls(struct task_struct *t, int tls, u32 addr) | |
258 | { | |
6612538c | 259 | struct user_desc ud = { |
1da177e4 LT |
260 | .base_addr = addr, |
261 | .limit = 0xfffff, | |
262 | .seg_32bit = 1, | |
263 | .limit_in_pages = 1, | |
264 | .useable = 1, | |
265 | }; | |
ade1af77 | 266 | struct desc_struct *desc = t->thread.tls_array; |
1da177e4 | 267 | desc += tls; |
80fbb69a | 268 | fill_ldt(desc, &ud); |
1da177e4 LT |
269 | } |
270 | ||
271 | static inline u32 read_32bit_tls(struct task_struct *t, int tls) | |
272 | { | |
91394eb0 | 273 | return get_desc_base(&t->thread.tls_array[tls]); |
1da177e4 LT |
274 | } |
275 | ||
276 | /* | |
277 | * This gets called before we allocate a new thread and copy | |
278 | * the current task into it. | |
279 | */ | |
280 | void prepare_to_copy(struct task_struct *tsk) | |
281 | { | |
282 | unlazy_fpu(tsk); | |
283 | } | |
284 | ||
6f2c55b8 | 285 | int copy_thread(unsigned long clone_flags, unsigned long sp, |
1da177e4 | 286 | unsigned long unused, |
7de08b4e | 287 | struct task_struct *p, struct pt_regs *regs) |
1da177e4 LT |
288 | { |
289 | int err; | |
7de08b4e | 290 | struct pt_regs *childregs; |
1da177e4 LT |
291 | struct task_struct *me = current; |
292 | ||
a88cde13 | 293 | childregs = ((struct pt_regs *) |
57eafdc2 | 294 | (THREAD_SIZE + task_stack_page(p))) - 1; |
1da177e4 LT |
295 | *childregs = *regs; |
296 | ||
65ea5b03 PA |
297 | childregs->ax = 0; |
298 | childregs->sp = sp; | |
299 | if (sp == ~0UL) | |
300 | childregs->sp = (unsigned long)childregs; | |
1da177e4 | 301 | |
faca6227 PA |
302 | p->thread.sp = (unsigned long) childregs; |
303 | p->thread.sp0 = (unsigned long) (childregs+1); | |
304 | p->thread.usersp = me->thread.usersp; | |
1da177e4 | 305 | |
e4f17c43 | 306 | set_tsk_thread_flag(p, TIF_FORK); |
1da177e4 LT |
307 | |
308 | p->thread.fs = me->thread.fs; | |
309 | p->thread.gs = me->thread.gs; | |
66cb5917 | 310 | p->thread.io_bitmap_ptr = NULL; |
1da177e4 | 311 | |
ada85708 JF |
312 | savesegment(gs, p->thread.gsindex); |
313 | savesegment(fs, p->thread.fsindex); | |
314 | savesegment(es, p->thread.es); | |
315 | savesegment(ds, p->thread.ds); | |
1da177e4 | 316 | |
66cb5917 P |
317 | err = -ENOMEM; |
318 | if (unlikely(test_tsk_thread_flag(me, TIF_DEBUG))) | |
319 | if (copy_thread_hw_breakpoint(me, p, clone_flags)) | |
320 | goto out; | |
321 | ||
d3a4f48d | 322 | if (unlikely(test_tsk_thread_flag(me, TIF_IO_BITMAP))) { |
1da177e4 LT |
323 | p->thread.io_bitmap_ptr = kmalloc(IO_BITMAP_BYTES, GFP_KERNEL); |
324 | if (!p->thread.io_bitmap_ptr) { | |
325 | p->thread.io_bitmap_max = 0; | |
326 | return -ENOMEM; | |
327 | } | |
a88cde13 AK |
328 | memcpy(p->thread.io_bitmap_ptr, me->thread.io_bitmap_ptr, |
329 | IO_BITMAP_BYTES); | |
d3a4f48d | 330 | set_tsk_thread_flag(p, TIF_IO_BITMAP); |
6612538c | 331 | } |
1da177e4 LT |
332 | |
333 | /* | |
334 | * Set a new TLS for the child thread? | |
335 | */ | |
336 | if (clone_flags & CLONE_SETTLS) { | |
337 | #ifdef CONFIG_IA32_EMULATION | |
338 | if (test_thread_flag(TIF_IA32)) | |
efd1ca52 | 339 | err = do_set_thread_area(p, -1, |
65ea5b03 | 340 | (struct user_desc __user *)childregs->si, 0); |
7de08b4e GP |
341 | else |
342 | #endif | |
343 | err = do_arch_prctl(p, ARCH_SET_FS, childregs->r8); | |
344 | if (err) | |
1da177e4 LT |
345 | goto out; |
346 | } | |
bf53de90 | 347 | |
2311f0de MM |
348 | clear_tsk_thread_flag(p, TIF_DS_AREA_MSR); |
349 | p->thread.ds_ctx = NULL; | |
bf53de90 MM |
350 | |
351 | clear_tsk_thread_flag(p, TIF_DEBUGCTLMSR); | |
352 | p->thread.debugctlmsr = 0; | |
353 | ||
1da177e4 LT |
354 | err = 0; |
355 | out: | |
356 | if (err && p->thread.io_bitmap_ptr) { | |
357 | kfree(p->thread.io_bitmap_ptr); | |
358 | p->thread.io_bitmap_max = 0; | |
359 | } | |
66cb5917 P |
360 | if (err) |
361 | flush_thread_hw_breakpoint(p); | |
362 | ||
1da177e4 LT |
363 | return err; |
364 | } | |
365 | ||
513ad84b IM |
366 | void |
367 | start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp) | |
368 | { | |
ada85708 JF |
369 | loadsegment(fs, 0); |
370 | loadsegment(es, 0); | |
371 | loadsegment(ds, 0); | |
513ad84b IM |
372 | load_gs_index(0); |
373 | regs->ip = new_ip; | |
374 | regs->sp = new_sp; | |
3d1e42a7 | 375 | percpu_write(old_rsp, new_sp); |
513ad84b IM |
376 | regs->cs = __USER_CS; |
377 | regs->ss = __USER_DS; | |
378 | regs->flags = 0x200; | |
379 | set_fs(USER_DS); | |
aa283f49 SS |
380 | /* |
381 | * Free the old FP and other extended state | |
382 | */ | |
383 | free_thread_xstate(current); | |
513ad84b IM |
384 | } |
385 | EXPORT_SYMBOL_GPL(start_thread); | |
386 | ||
1da177e4 LT |
387 | /* |
388 | * switch_to(x,y) should switch tasks from x to y. | |
389 | * | |
6612538c | 390 | * This could still be optimized: |
1da177e4 LT |
391 | * - fold all the options into a flag word and test it with a single test. |
392 | * - could test fs/gs bitsliced | |
099f318b AK |
393 | * |
394 | * Kprobes not supported here. Set the probe on schedule instead. | |
8b96f011 | 395 | * Function graph tracer not supported too. |
1da177e4 | 396 | */ |
8b96f011 | 397 | __notrace_funcgraph struct task_struct * |
a88cde13 | 398 | __switch_to(struct task_struct *prev_p, struct task_struct *next_p) |
1da177e4 | 399 | { |
87b935a0 JF |
400 | struct thread_struct *prev = &prev_p->thread; |
401 | struct thread_struct *next = &next_p->thread; | |
6612538c | 402 | int cpu = smp_processor_id(); |
1da177e4 | 403 | struct tss_struct *tss = &per_cpu(init_tss, cpu); |
478de5a9 | 404 | unsigned fsindex, gsindex; |
1da177e4 | 405 | |
e07e23e1 | 406 | /* we're going to use this soon, after a few expensive things */ |
7de08b4e | 407 | if (next_p->fpu_counter > 5) |
61c4628b | 408 | prefetch(next->xstate); |
e07e23e1 | 409 | |
1da177e4 LT |
410 | /* |
411 | * Reload esp0, LDT and the page table pointer: | |
412 | */ | |
7818a1e0 | 413 | load_sp0(tss, next); |
1da177e4 | 414 | |
7de08b4e | 415 | /* |
1da177e4 LT |
416 | * Switch DS and ES. |
417 | * This won't pick up thread selector changes, but I guess that is ok. | |
418 | */ | |
ada85708 | 419 | savesegment(es, prev->es); |
1da177e4 | 420 | if (unlikely(next->es | prev->es)) |
7de08b4e | 421 | loadsegment(es, next->es); |
ada85708 JF |
422 | |
423 | savesegment(ds, prev->ds); | |
1da177e4 LT |
424 | if (unlikely(next->ds | prev->ds)) |
425 | loadsegment(ds, next->ds); | |
426 | ||
478de5a9 JF |
427 | |
428 | /* We must save %fs and %gs before load_TLS() because | |
429 | * %fs and %gs may be cleared by load_TLS(). | |
430 | * | |
431 | * (e.g. xen_load_tls()) | |
432 | */ | |
433 | savesegment(fs, fsindex); | |
434 | savesegment(gs, gsindex); | |
435 | ||
1da177e4 LT |
436 | load_TLS(next, cpu); |
437 | ||
3fe0a63e JF |
438 | /* |
439 | * Leave lazy mode, flushing any hypercalls made here. | |
440 | * This must be done before restoring TLS segments so | |
441 | * the GDT and LDT are properly updated, and must be | |
442 | * done before math_state_restore, so the TS bit is up | |
443 | * to date. | |
444 | */ | |
445 | arch_leave_lazy_cpu_mode(); | |
446 | ||
7de08b4e | 447 | /* |
1da177e4 | 448 | * Switch FS and GS. |
87b935a0 JF |
449 | * |
450 | * Segment register != 0 always requires a reload. Also | |
451 | * reload when it has changed. When prev process used 64bit | |
452 | * base always reload to avoid an information leak. | |
1da177e4 | 453 | */ |
87b935a0 JF |
454 | if (unlikely(fsindex | next->fsindex | prev->fs)) { |
455 | loadsegment(fs, next->fsindex); | |
7de08b4e | 456 | /* |
87b935a0 JF |
457 | * Check if the user used a selector != 0; if yes |
458 | * clear 64bit base, since overloaded base is always | |
459 | * mapped to the Null selector | |
460 | */ | |
461 | if (fsindex) | |
7de08b4e | 462 | prev->fs = 0; |
1da177e4 | 463 | } |
87b935a0 JF |
464 | /* when next process has a 64bit base use it */ |
465 | if (next->fs) | |
466 | wrmsrl(MSR_FS_BASE, next->fs); | |
467 | prev->fsindex = fsindex; | |
468 | ||
469 | if (unlikely(gsindex | next->gsindex | prev->gs)) { | |
470 | load_gs_index(next->gsindex); | |
471 | if (gsindex) | |
7de08b4e | 472 | prev->gs = 0; |
1da177e4 | 473 | } |
87b935a0 JF |
474 | if (next->gs) |
475 | wrmsrl(MSR_KERNEL_GS_BASE, next->gs); | |
476 | prev->gsindex = gsindex; | |
1da177e4 | 477 | |
0a5ace2a AK |
478 | /* Must be after DS reload */ |
479 | unlazy_fpu(prev_p); | |
480 | ||
7de08b4e | 481 | /* |
45948d77 | 482 | * Switch the PDA and FPU contexts. |
1da177e4 | 483 | */ |
3d1e42a7 BG |
484 | prev->usersp = percpu_read(old_rsp); |
485 | percpu_write(old_rsp, next->usersp); | |
c6f5e0ac | 486 | percpu_write(current_task, next_p); |
18bd057b | 487 | |
9af45651 | 488 | percpu_write(kernel_stack, |
87b935a0 | 489 | (unsigned long)task_stack_page(next_p) + |
9af45651 | 490 | THREAD_SIZE - KERNEL_STACK_OFFSET); |
1da177e4 LT |
491 | |
492 | /* | |
d3a4f48d | 493 | * Now maybe reload the debug registers and handle I/O bitmaps |
1da177e4 | 494 | */ |
eee3af4a MM |
495 | if (unlikely(task_thread_info(next_p)->flags & _TIF_WORK_CTXSW_NEXT || |
496 | task_thread_info(prev_p)->flags & _TIF_WORK_CTXSW_PREV)) | |
d3a4f48d | 497 | __switch_to_xtra(prev_p, next_p, tss); |
1da177e4 | 498 | |
e07e23e1 AV |
499 | /* If the task has used fpu the last 5 timeslices, just do a full |
500 | * restore of the math state immediately to avoid the trap; the | |
501 | * chances of needing FPU soon are obviously high now | |
870568b3 SS |
502 | * |
503 | * tsk_used_math() checks prevent calling math_state_restore(), | |
504 | * which can sleep in the case of !tsk_used_math() | |
e07e23e1 | 505 | */ |
870568b3 | 506 | if (tsk_used_math(next_p) && next_p->fpu_counter > 5) |
e07e23e1 | 507 | math_state_restore(); |
66cb5917 P |
508 | /* |
509 | * There's a problem with moving the arch_install_thread_hw_breakpoint() | |
510 | * call before current is updated. Suppose a kernel breakpoint is | |
511 | * triggered in between the two, the hw-breakpoint handler will see that | |
512 | * the 'current' task does not have TIF_DEBUG flag set and will think it | |
513 | * is leftover from an old task (lazy switching) and will erase it. Then | |
514 | * until the next context switch, no user-breakpoints will be installed. | |
515 | * | |
516 | * The real problem is that it's impossible to update both current and | |
517 | * physical debug registers at the same instant, so there will always be | |
518 | * a window in which they disagree and a breakpoint might get triggered. | |
519 | * Since we use lazy switching, we are forced to assume that a | |
520 | * disagreement means that current is correct and the exception is due | |
521 | * to lazy debug register switching. | |
522 | */ | |
523 | if (unlikely(test_tsk_thread_flag(next_p, TIF_DEBUG))) | |
524 | arch_install_thread_hw_breakpoint(next_p); | |
525 | ||
1da177e4 LT |
526 | return prev_p; |
527 | } | |
528 | ||
529 | /* | |
530 | * sys_execve() executes a new program. | |
531 | */ | |
6612538c | 532 | asmlinkage |
1da177e4 | 533 | long sys_execve(char __user *name, char __user * __user *argv, |
5d119b2c | 534 | char __user * __user *envp, struct pt_regs *regs) |
1da177e4 LT |
535 | { |
536 | long error; | |
7de08b4e | 537 | char *filename; |
1da177e4 LT |
538 | |
539 | filename = getname(name); | |
540 | error = PTR_ERR(filename); | |
5d119b2c | 541 | if (IS_ERR(filename)) |
1da177e4 | 542 | return error; |
5d119b2c | 543 | error = do_execve(filename, argv, envp, regs); |
1da177e4 LT |
544 | putname(filename); |
545 | return error; | |
546 | } | |
547 | ||
548 | void set_personality_64bit(void) | |
549 | { | |
550 | /* inherit personality from parent */ | |
551 | ||
552 | /* Make sure to be in 64bit mode */ | |
6612538c | 553 | clear_thread_flag(TIF_IA32); |
1da177e4 LT |
554 | |
555 | /* TBD: overwrites user setup. Should have two bits. | |
556 | But 64bit processes have always behaved this way, | |
557 | so it's not too bad. The main problem is just that | |
6612538c | 558 | 32bit childs are affected again. */ |
1da177e4 LT |
559 | current->personality &= ~READ_IMPLIES_EXEC; |
560 | } | |
561 | ||
a88cde13 AK |
562 | asmlinkage long |
563 | sys_clone(unsigned long clone_flags, unsigned long newsp, | |
564 | void __user *parent_tid, void __user *child_tid, struct pt_regs *regs) | |
1da177e4 LT |
565 | { |
566 | if (!newsp) | |
65ea5b03 | 567 | newsp = regs->sp; |
1da177e4 LT |
568 | return do_fork(clone_flags, newsp, regs, 0, parent_tid, child_tid); |
569 | } | |
570 | ||
1da177e4 LT |
571 | unsigned long get_wchan(struct task_struct *p) |
572 | { | |
573 | unsigned long stack; | |
7de08b4e | 574 | u64 fp, ip; |
1da177e4 LT |
575 | int count = 0; |
576 | ||
7de08b4e GP |
577 | if (!p || p == current || p->state == TASK_RUNNING) |
578 | return 0; | |
57eafdc2 | 579 | stack = (unsigned long)task_stack_page(p); |
e1e23bb0 | 580 | if (p->thread.sp < stack || p->thread.sp >= stack+THREAD_SIZE) |
1da177e4 | 581 | return 0; |
faca6227 | 582 | fp = *(u64 *)(p->thread.sp); |
7de08b4e | 583 | do { |
a88cde13 | 584 | if (fp < (unsigned long)stack || |
e1e23bb0 | 585 | fp >= (unsigned long)stack+THREAD_SIZE) |
7de08b4e | 586 | return 0; |
65ea5b03 PA |
587 | ip = *(u64 *)(fp+8); |
588 | if (!in_sched_functions(ip)) | |
589 | return ip; | |
7de08b4e GP |
590 | fp = *(u64 *)fp; |
591 | } while (count++ < 16); | |
1da177e4 LT |
592 | return 0; |
593 | } | |
594 | ||
595 | long do_arch_prctl(struct task_struct *task, int code, unsigned long addr) | |
7de08b4e GP |
596 | { |
597 | int ret = 0; | |
1da177e4 LT |
598 | int doit = task == current; |
599 | int cpu; | |
600 | ||
7de08b4e | 601 | switch (code) { |
1da177e4 | 602 | case ARCH_SET_GS: |
84929801 | 603 | if (addr >= TASK_SIZE_OF(task)) |
7de08b4e | 604 | return -EPERM; |
1da177e4 | 605 | cpu = get_cpu(); |
7de08b4e | 606 | /* handle small bases via the GDT because that's faster to |
1da177e4 | 607 | switch. */ |
7de08b4e GP |
608 | if (addr <= 0xffffffff) { |
609 | set_32bit_tls(task, GS_TLS, addr); | |
610 | if (doit) { | |
1da177e4 | 611 | load_TLS(&task->thread, cpu); |
7de08b4e | 612 | load_gs_index(GS_TLS_SEL); |
1da177e4 | 613 | } |
7de08b4e | 614 | task->thread.gsindex = GS_TLS_SEL; |
1da177e4 | 615 | task->thread.gs = 0; |
7de08b4e | 616 | } else { |
1da177e4 LT |
617 | task->thread.gsindex = 0; |
618 | task->thread.gs = addr; | |
619 | if (doit) { | |
a88cde13 AK |
620 | load_gs_index(0); |
621 | ret = checking_wrmsrl(MSR_KERNEL_GS_BASE, addr); | |
7de08b4e | 622 | } |
1da177e4 LT |
623 | } |
624 | put_cpu(); | |
625 | break; | |
626 | case ARCH_SET_FS: | |
627 | /* Not strictly needed for fs, but do it for symmetry | |
628 | with gs */ | |
84929801 | 629 | if (addr >= TASK_SIZE_OF(task)) |
6612538c | 630 | return -EPERM; |
1da177e4 | 631 | cpu = get_cpu(); |
6612538c | 632 | /* handle small bases via the GDT because that's faster to |
1da177e4 | 633 | switch. */ |
6612538c | 634 | if (addr <= 0xffffffff) { |
1da177e4 | 635 | set_32bit_tls(task, FS_TLS, addr); |
6612538c HS |
636 | if (doit) { |
637 | load_TLS(&task->thread, cpu); | |
ada85708 | 638 | loadsegment(fs, FS_TLS_SEL); |
1da177e4 LT |
639 | } |
640 | task->thread.fsindex = FS_TLS_SEL; | |
641 | task->thread.fs = 0; | |
6612538c | 642 | } else { |
1da177e4 LT |
643 | task->thread.fsindex = 0; |
644 | task->thread.fs = addr; | |
645 | if (doit) { | |
646 | /* set the selector to 0 to not confuse | |
647 | __switch_to */ | |
ada85708 | 648 | loadsegment(fs, 0); |
a88cde13 | 649 | ret = checking_wrmsrl(MSR_FS_BASE, addr); |
1da177e4 LT |
650 | } |
651 | } | |
652 | put_cpu(); | |
653 | break; | |
6612538c HS |
654 | case ARCH_GET_FS: { |
655 | unsigned long base; | |
1da177e4 LT |
656 | if (task->thread.fsindex == FS_TLS_SEL) |
657 | base = read_32bit_tls(task, FS_TLS); | |
a88cde13 | 658 | else if (doit) |
1da177e4 | 659 | rdmsrl(MSR_FS_BASE, base); |
a88cde13 | 660 | else |
1da177e4 | 661 | base = task->thread.fs; |
6612538c HS |
662 | ret = put_user(base, (unsigned long __user *)addr); |
663 | break; | |
1da177e4 | 664 | } |
6612538c | 665 | case ARCH_GET_GS: { |
1da177e4 | 666 | unsigned long base; |
97c2803c | 667 | unsigned gsindex; |
1da177e4 LT |
668 | if (task->thread.gsindex == GS_TLS_SEL) |
669 | base = read_32bit_tls(task, GS_TLS); | |
97c2803c | 670 | else if (doit) { |
ada85708 | 671 | savesegment(gs, gsindex); |
97c2803c JB |
672 | if (gsindex) |
673 | rdmsrl(MSR_KERNEL_GS_BASE, base); | |
674 | else | |
675 | base = task->thread.gs; | |
7de08b4e | 676 | } else |
1da177e4 | 677 | base = task->thread.gs; |
6612538c | 678 | ret = put_user(base, (unsigned long __user *)addr); |
1da177e4 LT |
679 | break; |
680 | } | |
681 | ||
682 | default: | |
683 | ret = -EINVAL; | |
684 | break; | |
6612538c | 685 | } |
1da177e4 | 686 | |
6612538c HS |
687 | return ret; |
688 | } | |
1da177e4 LT |
689 | |
690 | long sys_arch_prctl(int code, unsigned long addr) | |
691 | { | |
692 | return do_arch_prctl(current, code, addr); | |
1da177e4 LT |
693 | } |
694 | ||
695 | unsigned long arch_align_stack(unsigned long sp) | |
696 | { | |
c16b63e0 | 697 | if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space) |
1da177e4 LT |
698 | sp -= get_random_int() % 8192; |
699 | return sp & ~0xf; | |
700 | } | |
c1d171a0 JK |
701 | |
702 | unsigned long arch_randomize_brk(struct mm_struct *mm) | |
703 | { | |
704 | unsigned long range_end = mm->brk + 0x02000000; | |
705 | return randomize_range(mm->brk, range_end, 0) ? : mm->brk; | |
706 | } |