Commit | Line | Data |
---|---|---|
14cf11af PM |
1 | /* |
2 | * arch/ppc/kernel/process.c | |
3 | * | |
4 | * Derived from "arch/i386/kernel/process.c" | |
5 | * Copyright (C) 1995 Linus Torvalds | |
6 | * | |
7 | * Updated and modified by Cort Dougan (cort@cs.nmt.edu) and | |
8 | * Paul Mackerras (paulus@cs.anu.edu.au) | |
9 | * | |
10 | * PowerPC version | |
11 | * Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org) | |
12 | * | |
13 | * This program is free software; you can redistribute it and/or | |
14 | * modify it under the terms of the GNU General Public License | |
15 | * as published by the Free Software Foundation; either version | |
16 | * 2 of the License, or (at your option) any later version. | |
17 | */ | |
18 | ||
19 | #include <linux/config.h> | |
20 | #include <linux/errno.h> | |
21 | #include <linux/sched.h> | |
22 | #include <linux/kernel.h> | |
23 | #include <linux/mm.h> | |
24 | #include <linux/smp.h> | |
25 | #include <linux/smp_lock.h> | |
26 | #include <linux/stddef.h> | |
27 | #include <linux/unistd.h> | |
28 | #include <linux/ptrace.h> | |
29 | #include <linux/slab.h> | |
30 | #include <linux/user.h> | |
31 | #include <linux/elf.h> | |
32 | #include <linux/init.h> | |
33 | #include <linux/prctl.h> | |
34 | #include <linux/init_task.h> | |
35 | #include <linux/module.h> | |
36 | #include <linux/kallsyms.h> | |
37 | #include <linux/mqueue.h> | |
38 | #include <linux/hardirq.h> | |
06d67d54 PM |
39 | #include <linux/utsname.h> |
40 | #include <linux/kprobes.h> | |
14cf11af PM |
41 | |
42 | #include <asm/pgtable.h> | |
43 | #include <asm/uaccess.h> | |
44 | #include <asm/system.h> | |
45 | #include <asm/io.h> | |
46 | #include <asm/processor.h> | |
47 | #include <asm/mmu.h> | |
48 | #include <asm/prom.h> | |
76032de8 | 49 | #include <asm/machdep.h> |
06d67d54 PM |
50 | #ifdef CONFIG_PPC64 |
51 | #include <asm/firmware.h> | |
06d67d54 PM |
52 | #include <asm/time.h> |
53 | #endif | |
14cf11af PM |
54 | |
55 | extern unsigned long _get_SP(void); | |
56 | ||
57 | #ifndef CONFIG_SMP | |
58 | struct task_struct *last_task_used_math = NULL; | |
59 | struct task_struct *last_task_used_altivec = NULL; | |
60 | struct task_struct *last_task_used_spe = NULL; | |
61 | #endif | |
62 | ||
14cf11af PM |
63 | /* |
64 | * Make sure the floating-point register state in the | |
65 | * the thread_struct is up to date for task tsk. | |
66 | */ | |
67 | void flush_fp_to_thread(struct task_struct *tsk) | |
68 | { | |
69 | if (tsk->thread.regs) { | |
70 | /* | |
71 | * We need to disable preemption here because if we didn't, | |
72 | * another process could get scheduled after the regs->msr | |
73 | * test but before we have finished saving the FP registers | |
74 | * to the thread_struct. That process could take over the | |
75 | * FPU, and then when we get scheduled again we would store | |
76 | * bogus values for the remaining FP registers. | |
77 | */ | |
78 | preempt_disable(); | |
79 | if (tsk->thread.regs->msr & MSR_FP) { | |
80 | #ifdef CONFIG_SMP | |
81 | /* | |
82 | * This should only ever be called for current or | |
83 | * for a stopped child process. Since we save away | |
84 | * the FP register state on context switch on SMP, | |
85 | * there is something wrong if a stopped child appears | |
86 | * to still have its FP state in the CPU registers. | |
87 | */ | |
88 | BUG_ON(tsk != current); | |
89 | #endif | |
90 | giveup_fpu(current); | |
91 | } | |
92 | preempt_enable(); | |
93 | } | |
94 | } | |
95 | ||
96 | void enable_kernel_fp(void) | |
97 | { | |
98 | WARN_ON(preemptible()); | |
99 | ||
100 | #ifdef CONFIG_SMP | |
101 | if (current->thread.regs && (current->thread.regs->msr & MSR_FP)) | |
102 | giveup_fpu(current); | |
103 | else | |
104 | giveup_fpu(NULL); /* just enables FP for kernel */ | |
105 | #else | |
106 | giveup_fpu(last_task_used_math); | |
107 | #endif /* CONFIG_SMP */ | |
108 | } | |
109 | EXPORT_SYMBOL(enable_kernel_fp); | |
110 | ||
111 | int dump_task_fpu(struct task_struct *tsk, elf_fpregset_t *fpregs) | |
112 | { | |
113 | if (!tsk->thread.regs) | |
114 | return 0; | |
115 | flush_fp_to_thread(current); | |
116 | ||
117 | memcpy(fpregs, &tsk->thread.fpr[0], sizeof(*fpregs)); | |
118 | ||
119 | return 1; | |
120 | } | |
121 | ||
122 | #ifdef CONFIG_ALTIVEC | |
123 | void enable_kernel_altivec(void) | |
124 | { | |
125 | WARN_ON(preemptible()); | |
126 | ||
127 | #ifdef CONFIG_SMP | |
128 | if (current->thread.regs && (current->thread.regs->msr & MSR_VEC)) | |
129 | giveup_altivec(current); | |
130 | else | |
131 | giveup_altivec(NULL); /* just enable AltiVec for kernel - force */ | |
132 | #else | |
133 | giveup_altivec(last_task_used_altivec); | |
134 | #endif /* CONFIG_SMP */ | |
135 | } | |
136 | EXPORT_SYMBOL(enable_kernel_altivec); | |
137 | ||
138 | /* | |
139 | * Make sure the VMX/Altivec register state in the | |
140 | * the thread_struct is up to date for task tsk. | |
141 | */ | |
142 | void flush_altivec_to_thread(struct task_struct *tsk) | |
143 | { | |
144 | if (tsk->thread.regs) { | |
145 | preempt_disable(); | |
146 | if (tsk->thread.regs->msr & MSR_VEC) { | |
147 | #ifdef CONFIG_SMP | |
148 | BUG_ON(tsk != current); | |
149 | #endif | |
150 | giveup_altivec(current); | |
151 | } | |
152 | preempt_enable(); | |
153 | } | |
154 | } | |
155 | ||
156 | int dump_task_altivec(struct pt_regs *regs, elf_vrregset_t *vrregs) | |
157 | { | |
158 | flush_altivec_to_thread(current); | |
159 | memcpy(vrregs, ¤t->thread.vr[0], sizeof(*vrregs)); | |
160 | return 1; | |
161 | } | |
162 | #endif /* CONFIG_ALTIVEC */ | |
163 | ||
164 | #ifdef CONFIG_SPE | |
165 | ||
166 | void enable_kernel_spe(void) | |
167 | { | |
168 | WARN_ON(preemptible()); | |
169 | ||
170 | #ifdef CONFIG_SMP | |
171 | if (current->thread.regs && (current->thread.regs->msr & MSR_SPE)) | |
172 | giveup_spe(current); | |
173 | else | |
174 | giveup_spe(NULL); /* just enable SPE for kernel - force */ | |
175 | #else | |
176 | giveup_spe(last_task_used_spe); | |
177 | #endif /* __SMP __ */ | |
178 | } | |
179 | EXPORT_SYMBOL(enable_kernel_spe); | |
180 | ||
181 | void flush_spe_to_thread(struct task_struct *tsk) | |
182 | { | |
183 | if (tsk->thread.regs) { | |
184 | preempt_disable(); | |
185 | if (tsk->thread.regs->msr & MSR_SPE) { | |
186 | #ifdef CONFIG_SMP | |
187 | BUG_ON(tsk != current); | |
188 | #endif | |
189 | giveup_spe(current); | |
190 | } | |
191 | preempt_enable(); | |
192 | } | |
193 | } | |
194 | ||
195 | int dump_spe(struct pt_regs *regs, elf_vrregset_t *evrregs) | |
196 | { | |
197 | flush_spe_to_thread(current); | |
198 | /* We copy u32 evr[32] + u64 acc + u32 spefscr -> 35 */ | |
199 | memcpy(evrregs, ¤t->thread.evr[0], sizeof(u32) * 35); | |
200 | return 1; | |
201 | } | |
202 | #endif /* CONFIG_SPE */ | |
203 | ||
48abec07 PM |
204 | /* |
205 | * If we are doing lazy switching of CPU state (FP, altivec or SPE), | |
206 | * and the current task has some state, discard it. | |
207 | */ | |
208 | static inline void discard_lazy_cpu_state(void) | |
209 | { | |
210 | #ifndef CONFIG_SMP | |
211 | preempt_disable(); | |
212 | if (last_task_used_math == current) | |
213 | last_task_used_math = NULL; | |
214 | #ifdef CONFIG_ALTIVEC | |
215 | if (last_task_used_altivec == current) | |
216 | last_task_used_altivec = NULL; | |
217 | #endif /* CONFIG_ALTIVEC */ | |
218 | #ifdef CONFIG_SPE | |
219 | if (last_task_used_spe == current) | |
220 | last_task_used_spe = NULL; | |
221 | #endif | |
222 | preempt_enable(); | |
223 | #endif /* CONFIG_SMP */ | |
224 | } | |
225 | ||
14cf11af PM |
226 | int set_dabr(unsigned long dabr) |
227 | { | |
cab0af98 ME |
228 | if (ppc_md.set_dabr) |
229 | return ppc_md.set_dabr(dabr); | |
14cf11af | 230 | |
cab0af98 ME |
231 | mtspr(SPRN_DABR, dabr); |
232 | return 0; | |
14cf11af PM |
233 | } |
234 | ||
06d67d54 PM |
235 | #ifdef CONFIG_PPC64 |
236 | DEFINE_PER_CPU(struct cpu_usage, cpu_usage_array); | |
14cf11af | 237 | static DEFINE_PER_CPU(unsigned long, current_dabr); |
06d67d54 | 238 | #endif |
14cf11af PM |
239 | |
240 | struct task_struct *__switch_to(struct task_struct *prev, | |
241 | struct task_struct *new) | |
242 | { | |
243 | struct thread_struct *new_thread, *old_thread; | |
244 | unsigned long flags; | |
245 | struct task_struct *last; | |
246 | ||
247 | #ifdef CONFIG_SMP | |
248 | /* avoid complexity of lazy save/restore of fpu | |
249 | * by just saving it every time we switch out if | |
250 | * this task used the fpu during the last quantum. | |
251 | * | |
252 | * If it tries to use the fpu again, it'll trap and | |
253 | * reload its fp regs. So we don't have to do a restore | |
254 | * every switch, just a save. | |
255 | * -- Cort | |
256 | */ | |
257 | if (prev->thread.regs && (prev->thread.regs->msr & MSR_FP)) | |
258 | giveup_fpu(prev); | |
259 | #ifdef CONFIG_ALTIVEC | |
260 | /* | |
261 | * If the previous thread used altivec in the last quantum | |
262 | * (thus changing altivec regs) then save them. | |
263 | * We used to check the VRSAVE register but not all apps | |
264 | * set it, so we don't rely on it now (and in fact we need | |
265 | * to save & restore VSCR even if VRSAVE == 0). -- paulus | |
266 | * | |
267 | * On SMP we always save/restore altivec regs just to avoid the | |
268 | * complexity of changing processors. | |
269 | * -- Cort | |
270 | */ | |
271 | if (prev->thread.regs && (prev->thread.regs->msr & MSR_VEC)) | |
272 | giveup_altivec(prev); | |
14cf11af PM |
273 | #endif /* CONFIG_ALTIVEC */ |
274 | #ifdef CONFIG_SPE | |
275 | /* | |
276 | * If the previous thread used spe in the last quantum | |
277 | * (thus changing spe regs) then save them. | |
278 | * | |
279 | * On SMP we always save/restore spe regs just to avoid the | |
280 | * complexity of changing processors. | |
281 | */ | |
282 | if ((prev->thread.regs && (prev->thread.regs->msr & MSR_SPE))) | |
283 | giveup_spe(prev); | |
c0c0d996 PM |
284 | #endif /* CONFIG_SPE */ |
285 | ||
286 | #else /* CONFIG_SMP */ | |
287 | #ifdef CONFIG_ALTIVEC | |
288 | /* Avoid the trap. On smp this this never happens since | |
289 | * we don't set last_task_used_altivec -- Cort | |
290 | */ | |
291 | if (new->thread.regs && last_task_used_altivec == new) | |
292 | new->thread.regs->msr |= MSR_VEC; | |
293 | #endif /* CONFIG_ALTIVEC */ | |
294 | #ifdef CONFIG_SPE | |
14cf11af PM |
295 | /* Avoid the trap. On smp this this never happens since |
296 | * we don't set last_task_used_spe | |
297 | */ | |
298 | if (new->thread.regs && last_task_used_spe == new) | |
299 | new->thread.regs->msr |= MSR_SPE; | |
300 | #endif /* CONFIG_SPE */ | |
c0c0d996 | 301 | |
14cf11af PM |
302 | #endif /* CONFIG_SMP */ |
303 | ||
304 | #ifdef CONFIG_PPC64 /* for now */ | |
305 | if (unlikely(__get_cpu_var(current_dabr) != new->thread.dabr)) { | |
306 | set_dabr(new->thread.dabr); | |
307 | __get_cpu_var(current_dabr) = new->thread.dabr; | |
308 | } | |
06d67d54 PM |
309 | |
310 | flush_tlb_pending(); | |
14cf11af PM |
311 | #endif |
312 | ||
313 | new_thread = &new->thread; | |
314 | old_thread = ¤t->thread; | |
06d67d54 PM |
315 | |
316 | #ifdef CONFIG_PPC64 | |
317 | /* | |
318 | * Collect processor utilization data per process | |
319 | */ | |
320 | if (firmware_has_feature(FW_FEATURE_SPLPAR)) { | |
321 | struct cpu_usage *cu = &__get_cpu_var(cpu_usage_array); | |
322 | long unsigned start_tb, current_tb; | |
323 | start_tb = old_thread->start_tb; | |
324 | cu->current_tb = current_tb = mfspr(SPRN_PURR); | |
325 | old_thread->accum_tb += (current_tb - start_tb); | |
326 | new_thread->start_tb = current_tb; | |
327 | } | |
328 | #endif | |
329 | ||
14cf11af PM |
330 | local_irq_save(flags); |
331 | last = _switch(old_thread, new_thread); | |
332 | ||
333 | local_irq_restore(flags); | |
334 | ||
335 | return last; | |
336 | } | |
337 | ||
06d67d54 PM |
338 | static int instructions_to_print = 16; |
339 | ||
340 | #ifdef CONFIG_PPC64 | |
341 | #define BAD_PC(pc) ((REGION_ID(pc) != KERNEL_REGION_ID) && \ | |
342 | (REGION_ID(pc) != VMALLOC_REGION_ID)) | |
343 | #else | |
344 | #define BAD_PC(pc) ((pc) < KERNELBASE) | |
345 | #endif | |
346 | ||
347 | static void show_instructions(struct pt_regs *regs) | |
348 | { | |
349 | int i; | |
350 | unsigned long pc = regs->nip - (instructions_to_print * 3 / 4 * | |
351 | sizeof(int)); | |
352 | ||
353 | printk("Instruction dump:"); | |
354 | ||
355 | for (i = 0; i < instructions_to_print; i++) { | |
356 | int instr; | |
357 | ||
358 | if (!(i % 8)) | |
359 | printk("\n"); | |
360 | ||
361 | if (BAD_PC(pc) || __get_user(instr, (unsigned int *)pc)) { | |
362 | printk("XXXXXXXX "); | |
363 | } else { | |
364 | if (regs->nip == pc) | |
365 | printk("<%08x> ", instr); | |
366 | else | |
367 | printk("%08x ", instr); | |
368 | } | |
369 | ||
370 | pc += sizeof(int); | |
371 | } | |
372 | ||
373 | printk("\n"); | |
374 | } | |
375 | ||
376 | static struct regbit { | |
377 | unsigned long bit; | |
378 | const char *name; | |
379 | } msr_bits[] = { | |
380 | {MSR_EE, "EE"}, | |
381 | {MSR_PR, "PR"}, | |
382 | {MSR_FP, "FP"}, | |
383 | {MSR_ME, "ME"}, | |
384 | {MSR_IR, "IR"}, | |
385 | {MSR_DR, "DR"}, | |
386 | {0, NULL} | |
387 | }; | |
388 | ||
389 | static void printbits(unsigned long val, struct regbit *bits) | |
390 | { | |
391 | const char *sep = ""; | |
392 | ||
393 | printk("<"); | |
394 | for (; bits->bit; ++bits) | |
395 | if (val & bits->bit) { | |
396 | printk("%s%s", sep, bits->name); | |
397 | sep = ","; | |
398 | } | |
399 | printk(">"); | |
400 | } | |
401 | ||
402 | #ifdef CONFIG_PPC64 | |
403 | #define REG "%016lX" | |
404 | #define REGS_PER_LINE 4 | |
405 | #define LAST_VOLATILE 13 | |
406 | #else | |
407 | #define REG "%08lX" | |
408 | #define REGS_PER_LINE 8 | |
409 | #define LAST_VOLATILE 12 | |
410 | #endif | |
411 | ||
14cf11af PM |
412 | void show_regs(struct pt_regs * regs) |
413 | { | |
414 | int i, trap; | |
415 | ||
06d67d54 PM |
416 | printk("NIP: "REG" LR: "REG" CTR: "REG"\n", |
417 | regs->nip, regs->link, regs->ctr); | |
418 | printk("REGS: %p TRAP: %04lx %s (%s)\n", | |
419 | regs, regs->trap, print_tainted(), system_utsname.release); | |
420 | printk("MSR: "REG" ", regs->msr); | |
421 | printbits(regs->msr, msr_bits); | |
422 | printk(" CR: %08lX XER: %08lX\n", regs->ccr, regs->xer); | |
14cf11af PM |
423 | trap = TRAP(regs); |
424 | if (trap == 0x300 || trap == 0x600) | |
06d67d54 PM |
425 | printk("DAR: "REG", DSISR: "REG"\n", regs->dar, regs->dsisr); |
426 | printk("TASK = %p[%d] '%s' THREAD: %p", | |
b5e2fc1c | 427 | current, current->pid, current->comm, task_thread_info(current)); |
14cf11af PM |
428 | |
429 | #ifdef CONFIG_SMP | |
430 | printk(" CPU: %d", smp_processor_id()); | |
431 | #endif /* CONFIG_SMP */ | |
432 | ||
433 | for (i = 0; i < 32; i++) { | |
06d67d54 | 434 | if ((i % REGS_PER_LINE) == 0) |
14cf11af | 435 | printk("\n" KERN_INFO "GPR%02d: ", i); |
06d67d54 PM |
436 | printk(REG " ", regs->gpr[i]); |
437 | if (i == LAST_VOLATILE && !FULL_REGS(regs)) | |
14cf11af PM |
438 | break; |
439 | } | |
440 | printk("\n"); | |
441 | #ifdef CONFIG_KALLSYMS | |
442 | /* | |
443 | * Lookup NIP late so we have the best change of getting the | |
444 | * above info out without failing | |
445 | */ | |
06d67d54 | 446 | printk("NIP ["REG"] ", regs->nip); |
14cf11af | 447 | print_symbol("%s\n", regs->nip); |
06d67d54 | 448 | printk("LR ["REG"] ", regs->link); |
14cf11af PM |
449 | print_symbol("%s\n", regs->link); |
450 | #endif | |
451 | show_stack(current, (unsigned long *) regs->gpr[1]); | |
06d67d54 PM |
452 | if (!user_mode(regs)) |
453 | show_instructions(regs); | |
14cf11af PM |
454 | } |
455 | ||
456 | void exit_thread(void) | |
457 | { | |
06d67d54 | 458 | kprobe_flush_task(current); |
48abec07 | 459 | discard_lazy_cpu_state(); |
14cf11af PM |
460 | } |
461 | ||
462 | void flush_thread(void) | |
463 | { | |
06d67d54 PM |
464 | #ifdef CONFIG_PPC64 |
465 | struct thread_info *t = current_thread_info(); | |
466 | ||
467 | if (t->flags & _TIF_ABI_PENDING) | |
468 | t->flags ^= (_TIF_ABI_PENDING | _TIF_32BIT); | |
469 | #endif | |
06d67d54 | 470 | |
48abec07 | 471 | discard_lazy_cpu_state(); |
14cf11af PM |
472 | |
473 | #ifdef CONFIG_PPC64 /* for now */ | |
474 | if (current->thread.dabr) { | |
475 | current->thread.dabr = 0; | |
476 | set_dabr(0); | |
477 | } | |
478 | #endif | |
479 | } | |
480 | ||
481 | void | |
482 | release_thread(struct task_struct *t) | |
483 | { | |
484 | } | |
485 | ||
486 | /* | |
487 | * This gets called before we allocate a new thread and copy | |
488 | * the current task into it. | |
489 | */ | |
490 | void prepare_to_copy(struct task_struct *tsk) | |
491 | { | |
492 | flush_fp_to_thread(current); | |
493 | flush_altivec_to_thread(current); | |
494 | flush_spe_to_thread(current); | |
495 | } | |
496 | ||
497 | /* | |
498 | * Copy a thread.. | |
499 | */ | |
06d67d54 PM |
500 | int copy_thread(int nr, unsigned long clone_flags, unsigned long usp, |
501 | unsigned long unused, struct task_struct *p, | |
502 | struct pt_regs *regs) | |
14cf11af PM |
503 | { |
504 | struct pt_regs *childregs, *kregs; | |
505 | extern void ret_from_fork(void); | |
506 | unsigned long sp = (unsigned long)p->thread_info + THREAD_SIZE; | |
14cf11af PM |
507 | |
508 | CHECK_FULL_REGS(regs); | |
509 | /* Copy registers */ | |
510 | sp -= sizeof(struct pt_regs); | |
511 | childregs = (struct pt_regs *) sp; | |
512 | *childregs = *regs; | |
513 | if ((childregs->msr & MSR_PR) == 0) { | |
514 | /* for kernel thread, set `current' and stackptr in new task */ | |
515 | childregs->gpr[1] = sp + sizeof(struct pt_regs); | |
06d67d54 | 516 | #ifdef CONFIG_PPC32 |
14cf11af | 517 | childregs->gpr[2] = (unsigned long) p; |
06d67d54 | 518 | #else |
b5e2fc1c | 519 | clear_tsk_thread_flag(p, TIF_32BIT); |
06d67d54 | 520 | #endif |
14cf11af PM |
521 | p->thread.regs = NULL; /* no user register state */ |
522 | } else { | |
523 | childregs->gpr[1] = usp; | |
524 | p->thread.regs = childregs; | |
06d67d54 PM |
525 | if (clone_flags & CLONE_SETTLS) { |
526 | #ifdef CONFIG_PPC64 | |
527 | if (!test_thread_flag(TIF_32BIT)) | |
528 | childregs->gpr[13] = childregs->gpr[6]; | |
529 | else | |
530 | #endif | |
531 | childregs->gpr[2] = childregs->gpr[6]; | |
532 | } | |
14cf11af PM |
533 | } |
534 | childregs->gpr[3] = 0; /* Result from fork() */ | |
535 | sp -= STACK_FRAME_OVERHEAD; | |
14cf11af PM |
536 | |
537 | /* | |
538 | * The way this works is that at some point in the future | |
539 | * some task will call _switch to switch to the new task. | |
540 | * That will pop off the stack frame created below and start | |
541 | * the new task running at ret_from_fork. The new task will | |
542 | * do some house keeping and then return from the fork or clone | |
543 | * system call, using the stack frame created above. | |
544 | */ | |
545 | sp -= sizeof(struct pt_regs); | |
546 | kregs = (struct pt_regs *) sp; | |
547 | sp -= STACK_FRAME_OVERHEAD; | |
548 | p->thread.ksp = sp; | |
14cf11af | 549 | |
06d67d54 PM |
550 | #ifdef CONFIG_PPC64 |
551 | if (cpu_has_feature(CPU_FTR_SLB)) { | |
552 | unsigned long sp_vsid = get_kernel_vsid(sp); | |
3c726f8d | 553 | unsigned long llp = mmu_psize_defs[mmu_linear_psize].sllp; |
06d67d54 PM |
554 | |
555 | sp_vsid <<= SLB_VSID_SHIFT; | |
3c726f8d | 556 | sp_vsid |= SLB_VSID_KERNEL | llp; |
06d67d54 PM |
557 | p->thread.ksp_vsid = sp_vsid; |
558 | } | |
559 | ||
560 | /* | |
561 | * The PPC64 ABI makes use of a TOC to contain function | |
562 | * pointers. The function (ret_from_except) is actually a pointer | |
563 | * to the TOC entry. The first entry is a pointer to the actual | |
564 | * function. | |
565 | */ | |
566 | kregs->nip = *((unsigned long *)ret_from_fork); | |
567 | #else | |
568 | kregs->nip = (unsigned long)ret_from_fork; | |
14cf11af | 569 | p->thread.last_syscall = -1; |
06d67d54 | 570 | #endif |
14cf11af PM |
571 | |
572 | return 0; | |
573 | } | |
574 | ||
575 | /* | |
576 | * Set up a thread for executing a new program | |
577 | */ | |
06d67d54 | 578 | void start_thread(struct pt_regs *regs, unsigned long start, unsigned long sp) |
14cf11af | 579 | { |
90eac727 ME |
580 | #ifdef CONFIG_PPC64 |
581 | unsigned long load_addr = regs->gpr[2]; /* saved by ELF_PLAT_INIT */ | |
582 | #endif | |
583 | ||
14cf11af | 584 | set_fs(USER_DS); |
06d67d54 PM |
585 | |
586 | /* | |
587 | * If we exec out of a kernel thread then thread.regs will not be | |
588 | * set. Do it now. | |
589 | */ | |
590 | if (!current->thread.regs) { | |
591 | unsigned long childregs = (unsigned long)current->thread_info + | |
592 | THREAD_SIZE; | |
593 | childregs -= sizeof(struct pt_regs); | |
594 | current->thread.regs = (struct pt_regs *)childregs; | |
595 | } | |
596 | ||
14cf11af PM |
597 | memset(regs->gpr, 0, sizeof(regs->gpr)); |
598 | regs->ctr = 0; | |
599 | regs->link = 0; | |
600 | regs->xer = 0; | |
601 | regs->ccr = 0; | |
14cf11af | 602 | regs->gpr[1] = sp; |
06d67d54 PM |
603 | |
604 | #ifdef CONFIG_PPC32 | |
605 | regs->mq = 0; | |
606 | regs->nip = start; | |
14cf11af | 607 | regs->msr = MSR_USER; |
06d67d54 | 608 | #else |
d4bf9a78 | 609 | if (!test_thread_flag(TIF_32BIT)) { |
90eac727 | 610 | unsigned long entry, toc; |
06d67d54 PM |
611 | |
612 | /* start is a relocated pointer to the function descriptor for | |
613 | * the elf _start routine. The first entry in the function | |
614 | * descriptor is the entry address of _start and the second | |
615 | * entry is the TOC value we need to use. | |
616 | */ | |
617 | __get_user(entry, (unsigned long __user *)start); | |
618 | __get_user(toc, (unsigned long __user *)start+1); | |
619 | ||
620 | /* Check whether the e_entry function descriptor entries | |
621 | * need to be relocated before we can use them. | |
622 | */ | |
623 | if (load_addr != 0) { | |
624 | entry += load_addr; | |
625 | toc += load_addr; | |
626 | } | |
627 | regs->nip = entry; | |
628 | regs->gpr[2] = toc; | |
629 | regs->msr = MSR_USER64; | |
d4bf9a78 SR |
630 | } else { |
631 | regs->nip = start; | |
632 | regs->gpr[2] = 0; | |
633 | regs->msr = MSR_USER32; | |
06d67d54 PM |
634 | } |
635 | #endif | |
636 | ||
48abec07 | 637 | discard_lazy_cpu_state(); |
14cf11af | 638 | memset(current->thread.fpr, 0, sizeof(current->thread.fpr)); |
25c8a78b | 639 | current->thread.fpscr.val = 0; |
14cf11af PM |
640 | #ifdef CONFIG_ALTIVEC |
641 | memset(current->thread.vr, 0, sizeof(current->thread.vr)); | |
642 | memset(¤t->thread.vscr, 0, sizeof(current->thread.vscr)); | |
06d67d54 | 643 | current->thread.vscr.u[3] = 0x00010000; /* Java mode disabled */ |
14cf11af PM |
644 | current->thread.vrsave = 0; |
645 | current->thread.used_vr = 0; | |
646 | #endif /* CONFIG_ALTIVEC */ | |
647 | #ifdef CONFIG_SPE | |
648 | memset(current->thread.evr, 0, sizeof(current->thread.evr)); | |
649 | current->thread.acc = 0; | |
650 | current->thread.spefscr = 0; | |
651 | current->thread.used_spe = 0; | |
652 | #endif /* CONFIG_SPE */ | |
653 | } | |
654 | ||
655 | #define PR_FP_ALL_EXCEPT (PR_FP_EXC_DIV | PR_FP_EXC_OVF | PR_FP_EXC_UND \ | |
656 | | PR_FP_EXC_RES | PR_FP_EXC_INV) | |
657 | ||
658 | int set_fpexc_mode(struct task_struct *tsk, unsigned int val) | |
659 | { | |
660 | struct pt_regs *regs = tsk->thread.regs; | |
661 | ||
662 | /* This is a bit hairy. If we are an SPE enabled processor | |
663 | * (have embedded fp) we store the IEEE exception enable flags in | |
664 | * fpexc_mode. fpexc_mode is also used for setting FP exception | |
665 | * mode (asyn, precise, disabled) for 'Classic' FP. */ | |
666 | if (val & PR_FP_EXC_SW_ENABLE) { | |
667 | #ifdef CONFIG_SPE | |
668 | tsk->thread.fpexc_mode = val & | |
669 | (PR_FP_EXC_SW_ENABLE | PR_FP_ALL_EXCEPT); | |
06d67d54 | 670 | return 0; |
14cf11af PM |
671 | #else |
672 | return -EINVAL; | |
673 | #endif | |
14cf11af | 674 | } |
06d67d54 PM |
675 | |
676 | /* on a CONFIG_SPE this does not hurt us. The bits that | |
677 | * __pack_fe01 use do not overlap with bits used for | |
678 | * PR_FP_EXC_SW_ENABLE. Additionally, the MSR[FE0,FE1] bits | |
679 | * on CONFIG_SPE implementations are reserved so writing to | |
680 | * them does not change anything */ | |
681 | if (val > PR_FP_EXC_PRECISE) | |
682 | return -EINVAL; | |
683 | tsk->thread.fpexc_mode = __pack_fe01(val); | |
684 | if (regs != NULL && (regs->msr & MSR_FP) != 0) | |
685 | regs->msr = (regs->msr & ~(MSR_FE0|MSR_FE1)) | |
686 | | tsk->thread.fpexc_mode; | |
14cf11af PM |
687 | return 0; |
688 | } | |
689 | ||
690 | int get_fpexc_mode(struct task_struct *tsk, unsigned long adr) | |
691 | { | |
692 | unsigned int val; | |
693 | ||
694 | if (tsk->thread.fpexc_mode & PR_FP_EXC_SW_ENABLE) | |
695 | #ifdef CONFIG_SPE | |
696 | val = tsk->thread.fpexc_mode; | |
697 | #else | |
698 | return -EINVAL; | |
699 | #endif | |
700 | else | |
701 | val = __unpack_fe01(tsk->thread.fpexc_mode); | |
702 | return put_user(val, (unsigned int __user *) adr); | |
703 | } | |
704 | ||
06d67d54 PM |
705 | #define TRUNC_PTR(x) ((typeof(x))(((unsigned long)(x)) & 0xffffffff)) |
706 | ||
14cf11af PM |
707 | int sys_clone(unsigned long clone_flags, unsigned long usp, |
708 | int __user *parent_tidp, void __user *child_threadptr, | |
709 | int __user *child_tidp, int p6, | |
710 | struct pt_regs *regs) | |
711 | { | |
712 | CHECK_FULL_REGS(regs); | |
713 | if (usp == 0) | |
714 | usp = regs->gpr[1]; /* stack pointer for child */ | |
06d67d54 PM |
715 | #ifdef CONFIG_PPC64 |
716 | if (test_thread_flag(TIF_32BIT)) { | |
717 | parent_tidp = TRUNC_PTR(parent_tidp); | |
718 | child_tidp = TRUNC_PTR(child_tidp); | |
719 | } | |
720 | #endif | |
14cf11af PM |
721 | return do_fork(clone_flags, usp, regs, 0, parent_tidp, child_tidp); |
722 | } | |
723 | ||
724 | int sys_fork(unsigned long p1, unsigned long p2, unsigned long p3, | |
725 | unsigned long p4, unsigned long p5, unsigned long p6, | |
726 | struct pt_regs *regs) | |
727 | { | |
728 | CHECK_FULL_REGS(regs); | |
729 | return do_fork(SIGCHLD, regs->gpr[1], regs, 0, NULL, NULL); | |
730 | } | |
731 | ||
732 | int sys_vfork(unsigned long p1, unsigned long p2, unsigned long p3, | |
733 | unsigned long p4, unsigned long p5, unsigned long p6, | |
734 | struct pt_regs *regs) | |
735 | { | |
736 | CHECK_FULL_REGS(regs); | |
737 | return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, regs->gpr[1], | |
738 | regs, 0, NULL, NULL); | |
739 | } | |
740 | ||
741 | int sys_execve(unsigned long a0, unsigned long a1, unsigned long a2, | |
742 | unsigned long a3, unsigned long a4, unsigned long a5, | |
743 | struct pt_regs *regs) | |
744 | { | |
745 | int error; | |
06d67d54 | 746 | char *filename; |
14cf11af PM |
747 | |
748 | filename = getname((char __user *) a0); | |
749 | error = PTR_ERR(filename); | |
750 | if (IS_ERR(filename)) | |
751 | goto out; | |
752 | flush_fp_to_thread(current); | |
753 | flush_altivec_to_thread(current); | |
754 | flush_spe_to_thread(current); | |
20c8c210 PM |
755 | error = do_execve(filename, (char __user * __user *) a1, |
756 | (char __user * __user *) a2, regs); | |
14cf11af PM |
757 | if (error == 0) { |
758 | task_lock(current); | |
759 | current->ptrace &= ~PT_DTRACE; | |
760 | task_unlock(current); | |
761 | } | |
762 | putname(filename); | |
763 | out: | |
764 | return error; | |
765 | } | |
766 | ||
767 | static int validate_sp(unsigned long sp, struct task_struct *p, | |
768 | unsigned long nbytes) | |
769 | { | |
770 | unsigned long stack_page = (unsigned long)p->thread_info; | |
771 | ||
772 | if (sp >= stack_page + sizeof(struct thread_struct) | |
773 | && sp <= stack_page + THREAD_SIZE - nbytes) | |
774 | return 1; | |
775 | ||
776 | #ifdef CONFIG_IRQSTACKS | |
777 | stack_page = (unsigned long) hardirq_ctx[task_cpu(p)]; | |
778 | if (sp >= stack_page + sizeof(struct thread_struct) | |
779 | && sp <= stack_page + THREAD_SIZE - nbytes) | |
780 | return 1; | |
781 | ||
782 | stack_page = (unsigned long) softirq_ctx[task_cpu(p)]; | |
783 | if (sp >= stack_page + sizeof(struct thread_struct) | |
784 | && sp <= stack_page + THREAD_SIZE - nbytes) | |
785 | return 1; | |
786 | #endif | |
787 | ||
788 | return 0; | |
789 | } | |
790 | ||
06d67d54 PM |
791 | #ifdef CONFIG_PPC64 |
792 | #define MIN_STACK_FRAME 112 /* same as STACK_FRAME_OVERHEAD, in fact */ | |
793 | #define FRAME_LR_SAVE 2 | |
794 | #define INT_FRAME_SIZE (sizeof(struct pt_regs) + STACK_FRAME_OVERHEAD + 288) | |
795 | #define REGS_MARKER 0x7265677368657265ul | |
796 | #define FRAME_MARKER 12 | |
797 | #else | |
798 | #define MIN_STACK_FRAME 16 | |
799 | #define FRAME_LR_SAVE 1 | |
800 | #define INT_FRAME_SIZE (sizeof(struct pt_regs) + STACK_FRAME_OVERHEAD) | |
801 | #define REGS_MARKER 0x72656773ul | |
802 | #define FRAME_MARKER 2 | |
14cf11af | 803 | #endif |
14cf11af PM |
804 | |
805 | unsigned long get_wchan(struct task_struct *p) | |
806 | { | |
807 | unsigned long ip, sp; | |
808 | int count = 0; | |
809 | ||
810 | if (!p || p == current || p->state == TASK_RUNNING) | |
811 | return 0; | |
812 | ||
813 | sp = p->thread.ksp; | |
06d67d54 | 814 | if (!validate_sp(sp, p, MIN_STACK_FRAME)) |
14cf11af PM |
815 | return 0; |
816 | ||
817 | do { | |
818 | sp = *(unsigned long *)sp; | |
06d67d54 | 819 | if (!validate_sp(sp, p, MIN_STACK_FRAME)) |
14cf11af PM |
820 | return 0; |
821 | if (count > 0) { | |
06d67d54 | 822 | ip = ((unsigned long *)sp)[FRAME_LR_SAVE]; |
14cf11af PM |
823 | if (!in_sched_functions(ip)) |
824 | return ip; | |
825 | } | |
826 | } while (count++ < 16); | |
827 | return 0; | |
828 | } | |
829 | EXPORT_SYMBOL(get_wchan); | |
06d67d54 PM |
830 | |
831 | static int kstack_depth_to_print = 64; | |
832 | ||
833 | void show_stack(struct task_struct *tsk, unsigned long *stack) | |
834 | { | |
835 | unsigned long sp, ip, lr, newsp; | |
836 | int count = 0; | |
837 | int firstframe = 1; | |
838 | ||
839 | sp = (unsigned long) stack; | |
840 | if (tsk == NULL) | |
841 | tsk = current; | |
842 | if (sp == 0) { | |
843 | if (tsk == current) | |
844 | asm("mr %0,1" : "=r" (sp)); | |
845 | else | |
846 | sp = tsk->thread.ksp; | |
847 | } | |
848 | ||
849 | lr = 0; | |
850 | printk("Call Trace:\n"); | |
851 | do { | |
852 | if (!validate_sp(sp, tsk, MIN_STACK_FRAME)) | |
853 | return; | |
854 | ||
855 | stack = (unsigned long *) sp; | |
856 | newsp = stack[0]; | |
857 | ip = stack[FRAME_LR_SAVE]; | |
858 | if (!firstframe || ip != lr) { | |
859 | printk("["REG"] ["REG"] ", sp, ip); | |
860 | print_symbol("%s", ip); | |
861 | if (firstframe) | |
862 | printk(" (unreliable)"); | |
863 | printk("\n"); | |
864 | } | |
865 | firstframe = 0; | |
866 | ||
867 | /* | |
868 | * See if this is an exception frame. | |
869 | * We look for the "regshere" marker in the current frame. | |
870 | */ | |
871 | if (validate_sp(sp, tsk, INT_FRAME_SIZE) | |
872 | && stack[FRAME_MARKER] == REGS_MARKER) { | |
873 | struct pt_regs *regs = (struct pt_regs *) | |
874 | (sp + STACK_FRAME_OVERHEAD); | |
875 | printk("--- Exception: %lx", regs->trap); | |
876 | print_symbol(" at %s\n", regs->nip); | |
877 | lr = regs->link; | |
878 | print_symbol(" LR = %s\n", lr); | |
879 | firstframe = 1; | |
880 | } | |
881 | ||
882 | sp = newsp; | |
883 | } while (count++ < kstack_depth_to_print); | |
884 | } | |
885 | ||
886 | void dump_stack(void) | |
887 | { | |
888 | show_stack(current, NULL); | |
889 | } | |
890 | EXPORT_SYMBOL(dump_stack); |