]>
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 | 28 | #include <linux/interrupt.h> |
6612538c | 29 | #include <linux/utsname.h> |
1da177e4 | 30 | #include <linux/delay.h> |
6612538c | 31 | #include <linux/module.h> |
1da177e4 | 32 | #include <linux/ptrace.h> |
95833c83 | 33 | #include <linux/notifier.h> |
c6fd91f0 | 34 | #include <linux/kprobes.h> |
1eeb66a1 | 35 | #include <linux/kdebug.h> |
02290683 | 36 | #include <linux/tick.h> |
529e25f6 | 37 | #include <linux/prctl.h> |
7de08b4e GP |
38 | #include <linux/uaccess.h> |
39 | #include <linux/io.h> | |
8b96f011 | 40 | #include <linux/ftrace.h> |
48ec4d95 | 41 | #include <linux/dmi.h> |
1da177e4 | 42 | |
1da177e4 LT |
43 | #include <asm/pgtable.h> |
44 | #include <asm/system.h> | |
1da177e4 LT |
45 | #include <asm/processor.h> |
46 | #include <asm/i387.h> | |
47 | #include <asm/mmu_context.h> | |
1da177e4 | 48 | #include <asm/prctl.h> |
1da177e4 LT |
49 | #include <asm/desc.h> |
50 | #include <asm/proto.h> | |
51 | #include <asm/ia32.h> | |
95833c83 | 52 | #include <asm/idle.h> |
bbc1f698 | 53 | #include <asm/syscalls.h> |
bf53de90 | 54 | #include <asm/ds.h> |
1da177e4 LT |
55 | |
56 | asmlinkage extern void ret_from_fork(void); | |
57 | ||
3d1e42a7 | 58 | DEFINE_PER_CPU(unsigned long, old_rsp); |
c2558e0e | 59 | static DEFINE_PER_CPU(unsigned char, is_idle); |
3d1e42a7 | 60 | |
1da177e4 LT |
61 | unsigned long kernel_thread_flags = CLONE_VM | CLONE_UNTRACED; |
62 | ||
e041c683 | 63 | static ATOMIC_NOTIFIER_HEAD(idle_notifier); |
95833c83 AK |
64 | |
65 | void idle_notifier_register(struct notifier_block *n) | |
66 | { | |
e041c683 | 67 | atomic_notifier_chain_register(&idle_notifier, n); |
95833c83 | 68 | } |
c7d87d79 VP |
69 | EXPORT_SYMBOL_GPL(idle_notifier_register); |
70 | ||
71 | void idle_notifier_unregister(struct notifier_block *n) | |
72 | { | |
73 | atomic_notifier_chain_unregister(&idle_notifier, n); | |
74 | } | |
75 | EXPORT_SYMBOL_GPL(idle_notifier_unregister); | |
95833c83 | 76 | |
95833c83 AK |
77 | void enter_idle(void) |
78 | { | |
c2558e0e | 79 | percpu_write(is_idle, 1); |
e041c683 | 80 | atomic_notifier_call_chain(&idle_notifier, IDLE_START, NULL); |
95833c83 AK |
81 | } |
82 | ||
83 | static void __exit_idle(void) | |
84 | { | |
c2558e0e | 85 | if (x86_test_and_clear_bit_percpu(0, is_idle) == 0) |
a15da49d | 86 | return; |
e041c683 | 87 | atomic_notifier_call_chain(&idle_notifier, IDLE_END, NULL); |
95833c83 AK |
88 | } |
89 | ||
90 | /* Called from interrupts to signify idle end */ | |
91 | void exit_idle(void) | |
92 | { | |
a15da49d AK |
93 | /* idle loop has pid 0 */ |
94 | if (current->pid) | |
95833c83 AK |
95 | return; |
96 | __exit_idle(); | |
97 | } | |
98 | ||
913da64b | 99 | #ifndef CONFIG_SMP |
76e4f660 AR |
100 | static inline void play_dead(void) |
101 | { | |
102 | BUG(); | |
103 | } | |
913da64b | 104 | #endif |
76e4f660 | 105 | |
1da177e4 LT |
106 | /* |
107 | * The idle thread. There's no useful work to be | |
108 | * done, so just try to conserve power and have a | |
109 | * low exit latency (ie sit in a loop waiting for | |
110 | * somebody to say that they'd like to reschedule) | |
111 | */ | |
b10db7f0 | 112 | void cpu_idle(void) |
1da177e4 | 113 | { |
495ab9c0 | 114 | current_thread_info()->status |= TS_POLLING; |
ce22bd92 | 115 | |
ce22bd92 | 116 | /* |
5c79d2a5 TH |
117 | * If we're the non-boot CPU, nothing set the stack canary up |
118 | * for us. CPU0 already has it initialized but no harm in | |
119 | * doing it again. This is a good place for updating it, as | |
120 | * we wont ever return from this function (so the invalid | |
121 | * canaries already on the stack wont ever trigger). | |
ce22bd92 | 122 | */ |
18aa8bb1 IM |
123 | boot_init_stack_canary(); |
124 | ||
1da177e4 LT |
125 | /* endless idle loop with no priority at all */ |
126 | while (1) { | |
b8f8c3cf | 127 | tick_nohz_stop_sched_tick(1); |
1da177e4 | 128 | while (!need_resched()) { |
1da177e4 | 129 | |
1da177e4 | 130 | rmb(); |
6ddd2a27 | 131 | |
76e4f660 AR |
132 | if (cpu_is_offline(smp_processor_id())) |
133 | play_dead(); | |
d331e739 VP |
134 | /* |
135 | * Idle routines should keep interrupts disabled | |
136 | * from here on, until they go to idle. | |
137 | * Otherwise, idle callbacks can misfire. | |
138 | */ | |
139 | local_irq_disable(); | |
95833c83 | 140 | enter_idle(); |
81d68a96 SR |
141 | /* Don't trace irqs off for idle */ |
142 | stop_critical_timings(); | |
6ddd2a27 | 143 | pm_idle(); |
81d68a96 | 144 | start_critical_timings(); |
a15da49d AK |
145 | /* In many cases the interrupt that ended idle |
146 | has already called exit_idle. But some idle | |
147 | loops can be woken up without interrupt. */ | |
95833c83 | 148 | __exit_idle(); |
1da177e4 LT |
149 | } |
150 | ||
02290683 | 151 | tick_nohz_restart_sched_tick(); |
5bfb5d69 | 152 | preempt_enable_no_resched(); |
1da177e4 | 153 | schedule(); |
5bfb5d69 | 154 | preempt_disable(); |
1da177e4 LT |
155 | } |
156 | } | |
157 | ||
6612538c | 158 | /* Prints also some state that isn't saved in the pt_regs */ |
e2ce07c8 | 159 | void __show_regs(struct pt_regs *regs, int all) |
1da177e4 LT |
160 | { |
161 | unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L, fs, gs, shadowgs; | |
bb1995d5 | 162 | unsigned long d0, d1, d2, d3, d6, d7; |
6612538c HS |
163 | unsigned int fsindex, gsindex; |
164 | unsigned int ds, cs, es; | |
48ec4d95 | 165 | const char *board; |
1da177e4 LT |
166 | |
167 | printk("\n"); | |
168 | print_modules(); | |
48ec4d95 KM |
169 | board = dmi_get_system_info(DMI_PRODUCT_NAME); |
170 | if (!board) | |
171 | board = ""; | |
172 | printk(KERN_INFO "Pid: %d, comm: %.20s %s %s %.*s %s\n", | |
9acf23c4 | 173 | current->pid, current->comm, print_tainted(), |
96b644bd SH |
174 | init_utsname()->release, |
175 | (int)strcspn(init_utsname()->version, " "), | |
48ec4d95 | 176 | init_utsname()->version, board); |
8092c654 | 177 | printk(KERN_INFO "RIP: %04lx:[<%016lx>] ", regs->cs & 0xffff, regs->ip); |
aafbd7eb | 178 | printk_address(regs->ip, 1); |
8092c654 GP |
179 | printk(KERN_INFO "RSP: %04lx:%016lx EFLAGS: %08lx\n", regs->ss, |
180 | regs->sp, regs->flags); | |
181 | printk(KERN_INFO "RAX: %016lx RBX: %016lx RCX: %016lx\n", | |
65ea5b03 | 182 | regs->ax, regs->bx, regs->cx); |
8092c654 | 183 | printk(KERN_INFO "RDX: %016lx RSI: %016lx RDI: %016lx\n", |
65ea5b03 | 184 | regs->dx, regs->si, regs->di); |
8092c654 | 185 | printk(KERN_INFO "RBP: %016lx R08: %016lx R09: %016lx\n", |
65ea5b03 | 186 | regs->bp, regs->r8, regs->r9); |
8092c654 | 187 | printk(KERN_INFO "R10: %016lx R11: %016lx R12: %016lx\n", |
7de08b4e | 188 | regs->r10, regs->r11, regs->r12); |
8092c654 | 189 | printk(KERN_INFO "R13: %016lx R14: %016lx R15: %016lx\n", |
7de08b4e | 190 | regs->r13, regs->r14, regs->r15); |
1da177e4 | 191 | |
7de08b4e GP |
192 | asm("movl %%ds,%0" : "=r" (ds)); |
193 | asm("movl %%cs,%0" : "=r" (cs)); | |
194 | asm("movl %%es,%0" : "=r" (es)); | |
1da177e4 LT |
195 | asm("movl %%fs,%0" : "=r" (fsindex)); |
196 | asm("movl %%gs,%0" : "=r" (gsindex)); | |
197 | ||
198 | rdmsrl(MSR_FS_BASE, fs); | |
7de08b4e GP |
199 | rdmsrl(MSR_GS_BASE, gs); |
200 | rdmsrl(MSR_KERNEL_GS_BASE, shadowgs); | |
1da177e4 | 201 | |
e2ce07c8 PE |
202 | if (!all) |
203 | return; | |
1da177e4 | 204 | |
f51c9452 GOC |
205 | cr0 = read_cr0(); |
206 | cr2 = read_cr2(); | |
207 | cr3 = read_cr3(); | |
208 | cr4 = read_cr4(); | |
1da177e4 | 209 | |
8092c654 | 210 | printk(KERN_INFO "FS: %016lx(%04x) GS:%016lx(%04x) knlGS:%016lx\n", |
7de08b4e | 211 | fs, fsindex, gs, gsindex, shadowgs); |
8092c654 GP |
212 | printk(KERN_INFO "CS: %04x DS: %04x ES: %04x CR0: %016lx\n", cs, ds, |
213 | es, cr0); | |
214 | printk(KERN_INFO "CR2: %016lx CR3: %016lx CR4: %016lx\n", cr2, cr3, | |
215 | cr4); | |
bb1995d5 AS |
216 | |
217 | get_debugreg(d0, 0); | |
218 | get_debugreg(d1, 1); | |
219 | get_debugreg(d2, 2); | |
8092c654 | 220 | printk(KERN_INFO "DR0: %016lx DR1: %016lx DR2: %016lx\n", d0, d1, d2); |
bb1995d5 AS |
221 | get_debugreg(d3, 3); |
222 | get_debugreg(d6, 6); | |
223 | get_debugreg(d7, 7); | |
8092c654 | 224 | printk(KERN_INFO "DR3: %016lx DR6: %016lx DR7: %016lx\n", d3, d6, d7); |
1da177e4 LT |
225 | } |
226 | ||
227 | void show_regs(struct pt_regs *regs) | |
228 | { | |
8092c654 | 229 | printk(KERN_INFO "CPU %d:", smp_processor_id()); |
e2ce07c8 | 230 | __show_regs(regs, 1); |
bc850d6b | 231 | show_trace(NULL, regs, (void *)(regs + 1), regs->bp); |
1da177e4 LT |
232 | } |
233 | ||
1da177e4 LT |
234 | void release_thread(struct task_struct *dead_task) |
235 | { | |
236 | if (dead_task->mm) { | |
237 | if (dead_task->mm->context.size) { | |
238 | printk("WARNING: dead process %8s still has LDT? <%p/%d>\n", | |
239 | dead_task->comm, | |
240 | dead_task->mm->context.ldt, | |
241 | dead_task->mm->context.size); | |
242 | BUG(); | |
243 | } | |
244 | } | |
245 | } | |
246 | ||
247 | static inline void set_32bit_tls(struct task_struct *t, int tls, u32 addr) | |
248 | { | |
6612538c | 249 | struct user_desc ud = { |
1da177e4 LT |
250 | .base_addr = addr, |
251 | .limit = 0xfffff, | |
252 | .seg_32bit = 1, | |
253 | .limit_in_pages = 1, | |
254 | .useable = 1, | |
255 | }; | |
ade1af77 | 256 | struct desc_struct *desc = t->thread.tls_array; |
1da177e4 | 257 | desc += tls; |
80fbb69a | 258 | fill_ldt(desc, &ud); |
1da177e4 LT |
259 | } |
260 | ||
261 | static inline u32 read_32bit_tls(struct task_struct *t, int tls) | |
262 | { | |
91394eb0 | 263 | return get_desc_base(&t->thread.tls_array[tls]); |
1da177e4 LT |
264 | } |
265 | ||
266 | /* | |
267 | * This gets called before we allocate a new thread and copy | |
268 | * the current task into it. | |
269 | */ | |
270 | void prepare_to_copy(struct task_struct *tsk) | |
271 | { | |
272 | unlazy_fpu(tsk); | |
273 | } | |
274 | ||
6f2c55b8 | 275 | int copy_thread(unsigned long clone_flags, unsigned long sp, |
1da177e4 | 276 | unsigned long unused, |
7de08b4e | 277 | struct task_struct *p, struct pt_regs *regs) |
1da177e4 LT |
278 | { |
279 | int err; | |
7de08b4e | 280 | struct pt_regs *childregs; |
1da177e4 LT |
281 | struct task_struct *me = current; |
282 | ||
a88cde13 | 283 | childregs = ((struct pt_regs *) |
57eafdc2 | 284 | (THREAD_SIZE + task_stack_page(p))) - 1; |
1da177e4 LT |
285 | *childregs = *regs; |
286 | ||
65ea5b03 PA |
287 | childregs->ax = 0; |
288 | childregs->sp = sp; | |
289 | if (sp == ~0UL) | |
290 | childregs->sp = (unsigned long)childregs; | |
1da177e4 | 291 | |
faca6227 PA |
292 | p->thread.sp = (unsigned long) childregs; |
293 | p->thread.sp0 = (unsigned long) (childregs+1); | |
294 | p->thread.usersp = me->thread.usersp; | |
1da177e4 | 295 | |
e4f17c43 | 296 | set_tsk_thread_flag(p, TIF_FORK); |
1da177e4 LT |
297 | |
298 | p->thread.fs = me->thread.fs; | |
299 | p->thread.gs = me->thread.gs; | |
300 | ||
ada85708 JF |
301 | savesegment(gs, p->thread.gsindex); |
302 | savesegment(fs, p->thread.fsindex); | |
303 | savesegment(es, p->thread.es); | |
304 | savesegment(ds, p->thread.ds); | |
1da177e4 | 305 | |
d3a4f48d | 306 | if (unlikely(test_tsk_thread_flag(me, TIF_IO_BITMAP))) { |
1da177e4 LT |
307 | p->thread.io_bitmap_ptr = kmalloc(IO_BITMAP_BYTES, GFP_KERNEL); |
308 | if (!p->thread.io_bitmap_ptr) { | |
309 | p->thread.io_bitmap_max = 0; | |
310 | return -ENOMEM; | |
311 | } | |
a88cde13 AK |
312 | memcpy(p->thread.io_bitmap_ptr, me->thread.io_bitmap_ptr, |
313 | IO_BITMAP_BYTES); | |
d3a4f48d | 314 | set_tsk_thread_flag(p, TIF_IO_BITMAP); |
6612538c | 315 | } |
1da177e4 LT |
316 | |
317 | /* | |
318 | * Set a new TLS for the child thread? | |
319 | */ | |
320 | if (clone_flags & CLONE_SETTLS) { | |
321 | #ifdef CONFIG_IA32_EMULATION | |
322 | if (test_thread_flag(TIF_IA32)) | |
efd1ca52 | 323 | err = do_set_thread_area(p, -1, |
65ea5b03 | 324 | (struct user_desc __user *)childregs->si, 0); |
7de08b4e GP |
325 | else |
326 | #endif | |
327 | err = do_arch_prctl(p, ARCH_SET_FS, childregs->r8); | |
328 | if (err) | |
1da177e4 LT |
329 | goto out; |
330 | } | |
bf53de90 | 331 | |
2311f0de MM |
332 | clear_tsk_thread_flag(p, TIF_DS_AREA_MSR); |
333 | p->thread.ds_ctx = NULL; | |
bf53de90 MM |
334 | |
335 | clear_tsk_thread_flag(p, TIF_DEBUGCTLMSR); | |
336 | p->thread.debugctlmsr = 0; | |
337 | ||
1da177e4 LT |
338 | err = 0; |
339 | out: | |
340 | if (err && p->thread.io_bitmap_ptr) { | |
341 | kfree(p->thread.io_bitmap_ptr); | |
342 | p->thread.io_bitmap_max = 0; | |
343 | } | |
344 | return err; | |
345 | } | |
346 | ||
513ad84b IM |
347 | void |
348 | start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp) | |
349 | { | |
ada85708 JF |
350 | loadsegment(fs, 0); |
351 | loadsegment(es, 0); | |
352 | loadsegment(ds, 0); | |
513ad84b IM |
353 | load_gs_index(0); |
354 | regs->ip = new_ip; | |
355 | regs->sp = new_sp; | |
3d1e42a7 | 356 | percpu_write(old_rsp, new_sp); |
513ad84b IM |
357 | regs->cs = __USER_CS; |
358 | regs->ss = __USER_DS; | |
a6f05a6a | 359 | regs->flags = X86_EFLAGS_IF; |
513ad84b | 360 | set_fs(USER_DS); |
aa283f49 SS |
361 | /* |
362 | * Free the old FP and other extended state | |
363 | */ | |
364 | free_thread_xstate(current); | |
513ad84b IM |
365 | } |
366 | EXPORT_SYMBOL_GPL(start_thread); | |
367 | ||
a6f05a6a PA |
368 | #ifdef CONFIG_IA32_EMULATION |
369 | void start_thread_ia32(struct pt_regs *regs, u32 new_ip, u32 new_sp) | |
370 | { | |
371 | loadsegment(fs, 0); | |
372 | loadsegment(ds, __USER32_DS); | |
373 | loadsegment(es, __USER32_DS); | |
374 | load_gs_index(0); | |
375 | regs->ip = new_ip; | |
376 | regs->sp = new_sp; | |
377 | percpu_write(old_rsp, new_sp); | |
378 | regs->cs = __USER32_CS; | |
379 | regs->ss = __USER32_DS; | |
380 | regs->flags = X86_EFLAGS_IF; | |
381 | set_fs(USER_DS); | |
382 | /* | |
383 | * Free the old FP and other extended state | |
384 | */ | |
385 | free_thread_xstate(current); | |
386 | } | |
387 | #endif | |
388 | ||
1da177e4 LT |
389 | /* |
390 | * switch_to(x,y) should switch tasks from x to y. | |
391 | * | |
6612538c | 392 | * This could still be optimized: |
1da177e4 LT |
393 | * - fold all the options into a flag word and test it with a single test. |
394 | * - could test fs/gs bitsliced | |
099f318b AK |
395 | * |
396 | * Kprobes not supported here. Set the probe on schedule instead. | |
8b96f011 | 397 | * Function graph tracer not supported too. |
1da177e4 | 398 | */ |
8b96f011 | 399 | __notrace_funcgraph struct task_struct * |
a88cde13 | 400 | __switch_to(struct task_struct *prev_p, struct task_struct *next_p) |
1da177e4 | 401 | { |
87b935a0 JF |
402 | struct thread_struct *prev = &prev_p->thread; |
403 | struct thread_struct *next = &next_p->thread; | |
6612538c | 404 | int cpu = smp_processor_id(); |
1da177e4 | 405 | struct tss_struct *tss = &per_cpu(init_tss, cpu); |
478de5a9 | 406 | unsigned fsindex, gsindex; |
17950c5b JF |
407 | bool preload_fpu; |
408 | ||
409 | /* | |
410 | * If the task has used fpu the last 5 timeslices, just do a full | |
411 | * restore of the math state immediately to avoid the trap; the | |
412 | * chances of needing FPU soon are obviously high now | |
413 | */ | |
414 | preload_fpu = tsk_used_math(next_p) && next_p->fpu_counter > 5; | |
1da177e4 | 415 | |
e07e23e1 | 416 | /* we're going to use this soon, after a few expensive things */ |
17950c5b | 417 | if (preload_fpu) |
61c4628b | 418 | prefetch(next->xstate); |
e07e23e1 | 419 | |
1da177e4 LT |
420 | /* |
421 | * Reload esp0, LDT and the page table pointer: | |
422 | */ | |
7818a1e0 | 423 | load_sp0(tss, next); |
1da177e4 | 424 | |
7de08b4e | 425 | /* |
1da177e4 LT |
426 | * Switch DS and ES. |
427 | * This won't pick up thread selector changes, but I guess that is ok. | |
428 | */ | |
ada85708 | 429 | savesegment(es, prev->es); |
1da177e4 | 430 | if (unlikely(next->es | prev->es)) |
7de08b4e | 431 | loadsegment(es, next->es); |
ada85708 JF |
432 | |
433 | savesegment(ds, prev->ds); | |
1da177e4 LT |
434 | if (unlikely(next->ds | prev->ds)) |
435 | loadsegment(ds, next->ds); | |
436 | ||
478de5a9 JF |
437 | |
438 | /* We must save %fs and %gs before load_TLS() because | |
439 | * %fs and %gs may be cleared by load_TLS(). | |
440 | * | |
441 | * (e.g. xen_load_tls()) | |
442 | */ | |
443 | savesegment(fs, fsindex); | |
444 | savesegment(gs, gsindex); | |
445 | ||
1da177e4 LT |
446 | load_TLS(next, cpu); |
447 | ||
16d9dbf0 JF |
448 | /* Must be after DS reload */ |
449 | unlazy_fpu(prev_p); | |
450 | ||
17950c5b JF |
451 | /* Make sure cpu is ready for new context */ |
452 | if (preload_fpu) | |
453 | clts(); | |
454 | ||
3fe0a63e JF |
455 | /* |
456 | * Leave lazy mode, flushing any hypercalls made here. | |
457 | * This must be done before restoring TLS segments so | |
458 | * the GDT and LDT are properly updated, and must be | |
459 | * done before math_state_restore, so the TS bit is up | |
460 | * to date. | |
461 | */ | |
224101ed | 462 | arch_end_context_switch(next_p); |
3fe0a63e | 463 | |
7de08b4e | 464 | /* |
1da177e4 | 465 | * Switch FS and GS. |
87b935a0 JF |
466 | * |
467 | * Segment register != 0 always requires a reload. Also | |
468 | * reload when it has changed. When prev process used 64bit | |
469 | * base always reload to avoid an information leak. | |
1da177e4 | 470 | */ |
87b935a0 JF |
471 | if (unlikely(fsindex | next->fsindex | prev->fs)) { |
472 | loadsegment(fs, next->fsindex); | |
7de08b4e | 473 | /* |
87b935a0 JF |
474 | * Check if the user used a selector != 0; if yes |
475 | * clear 64bit base, since overloaded base is always | |
476 | * mapped to the Null selector | |
477 | */ | |
478 | if (fsindex) | |
7de08b4e | 479 | prev->fs = 0; |
1da177e4 | 480 | } |
87b935a0 JF |
481 | /* when next process has a 64bit base use it */ |
482 | if (next->fs) | |
483 | wrmsrl(MSR_FS_BASE, next->fs); | |
484 | prev->fsindex = fsindex; | |
485 | ||
486 | if (unlikely(gsindex | next->gsindex | prev->gs)) { | |
487 | load_gs_index(next->gsindex); | |
488 | if (gsindex) | |
7de08b4e | 489 | prev->gs = 0; |
1da177e4 | 490 | } |
87b935a0 JF |
491 | if (next->gs) |
492 | wrmsrl(MSR_KERNEL_GS_BASE, next->gs); | |
493 | prev->gsindex = gsindex; | |
1da177e4 | 494 | |
7de08b4e | 495 | /* |
45948d77 | 496 | * Switch the PDA and FPU contexts. |
1da177e4 | 497 | */ |
3d1e42a7 BG |
498 | prev->usersp = percpu_read(old_rsp); |
499 | percpu_write(old_rsp, next->usersp); | |
c6f5e0ac | 500 | percpu_write(current_task, next_p); |
18bd057b | 501 | |
9af45651 | 502 | percpu_write(kernel_stack, |
87b935a0 | 503 | (unsigned long)task_stack_page(next_p) + |
9af45651 | 504 | THREAD_SIZE - KERNEL_STACK_OFFSET); |
1da177e4 LT |
505 | |
506 | /* | |
d3a4f48d | 507 | * Now maybe reload the debug registers and handle I/O bitmaps |
1da177e4 | 508 | */ |
eee3af4a MM |
509 | if (unlikely(task_thread_info(next_p)->flags & _TIF_WORK_CTXSW_NEXT || |
510 | task_thread_info(prev_p)->flags & _TIF_WORK_CTXSW_PREV)) | |
d3a4f48d | 511 | __switch_to_xtra(prev_p, next_p, tss); |
1da177e4 | 512 | |
17950c5b JF |
513 | /* |
514 | * Preload the FPU context, now that we've determined that the | |
515 | * task is likely to be using it. | |
e07e23e1 | 516 | */ |
17950c5b JF |
517 | if (preload_fpu) |
518 | __math_state_restore(); | |
1da177e4 LT |
519 | return prev_p; |
520 | } | |
521 | ||
522 | /* | |
523 | * sys_execve() executes a new program. | |
524 | */ | |
6612538c | 525 | asmlinkage |
1da177e4 | 526 | long sys_execve(char __user *name, char __user * __user *argv, |
5d119b2c | 527 | char __user * __user *envp, struct pt_regs *regs) |
1da177e4 LT |
528 | { |
529 | long error; | |
7de08b4e | 530 | char *filename; |
1da177e4 LT |
531 | |
532 | filename = getname(name); | |
533 | error = PTR_ERR(filename); | |
5d119b2c | 534 | if (IS_ERR(filename)) |
1da177e4 | 535 | return error; |
5d119b2c | 536 | error = do_execve(filename, argv, envp, regs); |
1da177e4 LT |
537 | putname(filename); |
538 | return error; | |
539 | } | |
540 | ||
541 | void set_personality_64bit(void) | |
542 | { | |
543 | /* inherit personality from parent */ | |
544 | ||
545 | /* Make sure to be in 64bit mode */ | |
6612538c | 546 | clear_thread_flag(TIF_IA32); |
1da177e4 LT |
547 | |
548 | /* TBD: overwrites user setup. Should have two bits. | |
549 | But 64bit processes have always behaved this way, | |
550 | so it's not too bad. The main problem is just that | |
6612538c | 551 | 32bit childs are affected again. */ |
1da177e4 LT |
552 | current->personality &= ~READ_IMPLIES_EXEC; |
553 | } | |
554 | ||
a88cde13 AK |
555 | asmlinkage long |
556 | sys_clone(unsigned long clone_flags, unsigned long newsp, | |
557 | void __user *parent_tid, void __user *child_tid, struct pt_regs *regs) | |
1da177e4 LT |
558 | { |
559 | if (!newsp) | |
65ea5b03 | 560 | newsp = regs->sp; |
1da177e4 LT |
561 | return do_fork(clone_flags, newsp, regs, 0, parent_tid, child_tid); |
562 | } | |
563 | ||
1da177e4 LT |
564 | unsigned long get_wchan(struct task_struct *p) |
565 | { | |
566 | unsigned long stack; | |
7de08b4e | 567 | u64 fp, ip; |
1da177e4 LT |
568 | int count = 0; |
569 | ||
7de08b4e GP |
570 | if (!p || p == current || p->state == TASK_RUNNING) |
571 | return 0; | |
57eafdc2 | 572 | stack = (unsigned long)task_stack_page(p); |
e1e23bb0 | 573 | if (p->thread.sp < stack || p->thread.sp >= stack+THREAD_SIZE) |
1da177e4 | 574 | return 0; |
faca6227 | 575 | fp = *(u64 *)(p->thread.sp); |
7de08b4e | 576 | do { |
a88cde13 | 577 | if (fp < (unsigned long)stack || |
e1e23bb0 | 578 | fp >= (unsigned long)stack+THREAD_SIZE) |
7de08b4e | 579 | return 0; |
65ea5b03 PA |
580 | ip = *(u64 *)(fp+8); |
581 | if (!in_sched_functions(ip)) | |
582 | return ip; | |
7de08b4e GP |
583 | fp = *(u64 *)fp; |
584 | } while (count++ < 16); | |
1da177e4 LT |
585 | return 0; |
586 | } | |
587 | ||
588 | long do_arch_prctl(struct task_struct *task, int code, unsigned long addr) | |
7de08b4e GP |
589 | { |
590 | int ret = 0; | |
1da177e4 LT |
591 | int doit = task == current; |
592 | int cpu; | |
593 | ||
7de08b4e | 594 | switch (code) { |
1da177e4 | 595 | case ARCH_SET_GS: |
84929801 | 596 | if (addr >= TASK_SIZE_OF(task)) |
7de08b4e | 597 | return -EPERM; |
1da177e4 | 598 | cpu = get_cpu(); |
7de08b4e | 599 | /* handle small bases via the GDT because that's faster to |
1da177e4 | 600 | switch. */ |
7de08b4e GP |
601 | if (addr <= 0xffffffff) { |
602 | set_32bit_tls(task, GS_TLS, addr); | |
603 | if (doit) { | |
1da177e4 | 604 | load_TLS(&task->thread, cpu); |
7de08b4e | 605 | load_gs_index(GS_TLS_SEL); |
1da177e4 | 606 | } |
7de08b4e | 607 | task->thread.gsindex = GS_TLS_SEL; |
1da177e4 | 608 | task->thread.gs = 0; |
7de08b4e | 609 | } else { |
1da177e4 LT |
610 | task->thread.gsindex = 0; |
611 | task->thread.gs = addr; | |
612 | if (doit) { | |
a88cde13 AK |
613 | load_gs_index(0); |
614 | ret = checking_wrmsrl(MSR_KERNEL_GS_BASE, addr); | |
7de08b4e | 615 | } |
1da177e4 LT |
616 | } |
617 | put_cpu(); | |
618 | break; | |
619 | case ARCH_SET_FS: | |
620 | /* Not strictly needed for fs, but do it for symmetry | |
621 | with gs */ | |
84929801 | 622 | if (addr >= TASK_SIZE_OF(task)) |
6612538c | 623 | return -EPERM; |
1da177e4 | 624 | cpu = get_cpu(); |
6612538c | 625 | /* handle small bases via the GDT because that's faster to |
1da177e4 | 626 | switch. */ |
6612538c | 627 | if (addr <= 0xffffffff) { |
1da177e4 | 628 | set_32bit_tls(task, FS_TLS, addr); |
6612538c HS |
629 | if (doit) { |
630 | load_TLS(&task->thread, cpu); | |
ada85708 | 631 | loadsegment(fs, FS_TLS_SEL); |
1da177e4 LT |
632 | } |
633 | task->thread.fsindex = FS_TLS_SEL; | |
634 | task->thread.fs = 0; | |
6612538c | 635 | } else { |
1da177e4 LT |
636 | task->thread.fsindex = 0; |
637 | task->thread.fs = addr; | |
638 | if (doit) { | |
639 | /* set the selector to 0 to not confuse | |
640 | __switch_to */ | |
ada85708 | 641 | loadsegment(fs, 0); |
a88cde13 | 642 | ret = checking_wrmsrl(MSR_FS_BASE, addr); |
1da177e4 LT |
643 | } |
644 | } | |
645 | put_cpu(); | |
646 | break; | |
6612538c HS |
647 | case ARCH_GET_FS: { |
648 | unsigned long base; | |
1da177e4 LT |
649 | if (task->thread.fsindex == FS_TLS_SEL) |
650 | base = read_32bit_tls(task, FS_TLS); | |
a88cde13 | 651 | else if (doit) |
1da177e4 | 652 | rdmsrl(MSR_FS_BASE, base); |
a88cde13 | 653 | else |
1da177e4 | 654 | base = task->thread.fs; |
6612538c HS |
655 | ret = put_user(base, (unsigned long __user *)addr); |
656 | break; | |
1da177e4 | 657 | } |
6612538c | 658 | case ARCH_GET_GS: { |
1da177e4 | 659 | unsigned long base; |
97c2803c | 660 | unsigned gsindex; |
1da177e4 LT |
661 | if (task->thread.gsindex == GS_TLS_SEL) |
662 | base = read_32bit_tls(task, GS_TLS); | |
97c2803c | 663 | else if (doit) { |
ada85708 | 664 | savesegment(gs, gsindex); |
97c2803c JB |
665 | if (gsindex) |
666 | rdmsrl(MSR_KERNEL_GS_BASE, base); | |
667 | else | |
668 | base = task->thread.gs; | |
7de08b4e | 669 | } else |
1da177e4 | 670 | base = task->thread.gs; |
6612538c | 671 | ret = put_user(base, (unsigned long __user *)addr); |
1da177e4 LT |
672 | break; |
673 | } | |
674 | ||
675 | default: | |
676 | ret = -EINVAL; | |
677 | break; | |
6612538c | 678 | } |
1da177e4 | 679 | |
6612538c HS |
680 | return ret; |
681 | } | |
1da177e4 LT |
682 | |
683 | long sys_arch_prctl(int code, unsigned long addr) | |
684 | { | |
685 | return do_arch_prctl(current, code, addr); | |
1da177e4 LT |
686 | } |
687 |