arch/mips/kernel/process.c

   1 /*
   2  * This file is subject to the terms and conditions of the GNU General Public
   3  * License.  See the file "COPYING" in the main directory of this archive
   4  * for more details.
   5  *
   6  * Copyright (C) 1994 - 1999, 2000 by Ralf Baechle and others.
   7  * Copyright (C) 2005, 2006 by Ralf Baechle (ralf@linux-mips.org)
   8  * Copyright (C) 1999, 2000 Silicon Graphics, Inc.
   9  * Copyright (C) 2004 Thiemo Seufer
  10  */
  11 #include <linux/errno.h>
  12 #include <linux/sched.h>
  13 #include <linux/tick.h>
  14 #include <linux/kernel.h>
  15 #include <linux/mm.h>
  16 #include <linux/stddef.h>
  17 #include <linux/unistd.h>
  18 #include <linux/export.h>
  19 #include <linux/ptrace.h>
  20 #include <linux/mman.h>
  21 #include <linux/personality.h>
  22 #include <linux/sys.h>
  23 #include <linux/user.h>
  24 #include <linux/init.h>
  25 #include <linux/completion.h>
  26 #include <linux/kallsyms.h>
  27 #include <linux/random.h>
  28
  29 #include <asm/asm.h>
  30 #include <asm/bootinfo.h>
  31 #include <asm/cpu.h>
  32 #include <asm/dsp.h>
  33 #include <asm/fpu.h>
  34 #include <asm/pgtable.h>
  35 #include <asm/system.h>
  36 #include <asm/mipsregs.h>
  37 #include <asm/processor.h>
  38 #include <asm/uaccess.h>
  39 #include <asm/io.h>
  40 #include <asm/elf.h>
  41 #include <asm/isadep.h>
  42 #include <asm/inst.h>
  43 #include <asm/stacktrace.h>
  44
  45 /*
  46  * The idle thread. There's no useful work to be done, so just try to conserve
  47  * power and have a low exit latency (ie sit in a loop waiting for somebody to
  48  * say that they'd like to reschedule)
  49  */
  50 void __noreturn cpu_idle(void)
  51 {
  52         int cpu;
  53
  54         /* CPU is going idle. */
  55         cpu = smp_processor_id();
  56
  57         /* endless idle loop with no priority at all */
  58         while (1) {
  59                 tick_nohz_idle_enter();
  60                 rcu_idle_enter();
  61                 while (!need_resched() && cpu_online(cpu)) {
  62 #ifdef CONFIG_MIPS_MT_SMTC
  63                         extern void smtc_idle_loop_hook(void);
  64
  65                         smtc_idle_loop_hook();
  66 #endif
  67
  68                         if (cpu_wait) {
  69                                 /* Don't trace irqs off for idle */
  70                                 stop_critical_timings();
  71                                 (*cpu_wait)();
  72                                 start_critical_timings();
  73                         }
  74                 }
  75 #ifdef CONFIG_HOTPLUG_CPU
  76                 if (!cpu_online(cpu) && !cpu_isset(cpu, cpu_callin_map) &&
  77                     (system_state == SYSTEM_RUNNING ||
  78                      system_state == SYSTEM_BOOTING))
  79                         play_dead();
  80 #endif
  81                 rcu_idle_exit();
  82                 tick_nohz_idle_exit();
  83                 schedule_preempt_disabled();
  84         }
  85 }
  86
  87 asmlinkage void ret_from_fork(void);
  88
  89 void start_thread(struct pt_regs * regs, unsigned long pc, unsigned long sp)
  90 {
  91         unsigned long status;
  92
  93         /* New thread loses kernel privileges. */
  94         status = regs->cp0_status & ~(ST0_CU0|ST0_CU1|ST0_FR|KU_MASK);
  95 #ifdef CONFIG_64BIT
  96         status |= test_thread_flag(TIF_32BIT_REGS) ? 0 : ST0_FR;
  97 #endif
  98         status |= KU_USER;
  99         regs->cp0_status = status;
 100         clear_used_math();
 101         clear_fpu_owner();
 102         if (cpu_has_dsp)
 103                 __init_dsp();
 104         regs->cp0_epc = pc;
 105         regs->regs[29] = sp;
 106 }
 107
 108 void exit_thread(void)
 109 {
 110 }
 111
 112 void flush_thread(void)
 113 {
 114 }
 115
 116 int copy_thread(unsigned long clone_flags, unsigned long usp,
 117         unsigned long unused, struct task_struct *p, struct pt_regs *regs)
 118 {
 119         struct thread_info *ti = task_thread_info(p);
 120         struct pt_regs *childregs;
 121         unsigned long childksp;
 122         p->set_child_tid = p->clear_child_tid = NULL;
 123
 124         childksp = (unsigned long)task_stack_page(p) + THREAD_SIZE - 32;
 125
 126         preempt_disable();
 127
 128         if (is_fpu_owner())
 129                 save_fp(p);
 130
 131         if (cpu_has_dsp)
 132                 save_dsp(p);
 133
 134         preempt_enable();
 135
 136         /* set up new TSS. */
 137         childregs = (struct pt_regs *) childksp - 1;
 138         /*  Put the stack after the struct pt_regs.  */
 139         childksp = (unsigned long) childregs;
 140         *childregs = *regs;
 141         childregs->regs[7] = 0; /* Clear error flag */
 142
 143         childregs->regs[2] = 0; /* Child gets zero as return value */
 144
 145         if (childregs->cp0_status & ST0_CU0) {
 146                 childregs->regs[28] = (unsigned long) ti;
 147                 childregs->regs[29] = childksp;
 148                 ti->addr_limit = KERNEL_DS;
 149         } else {
 150                 childregs->regs[29] = usp;
 151                 ti->addr_limit = USER_DS;
 152         }
 153         p->thread.reg29 = (unsigned long) childregs;
 154         p->thread.reg31 = (unsigned long) ret_from_fork;
 155
 156         /*
 157          * New tasks lose permission to use the fpu. This accelerates context
 158          * switching for most programs since they don't use the fpu.
 159          */
 160         p->thread.cp0_status = read_c0_status() & ~(ST0_CU2|ST0_CU1);
 161         childregs->cp0_status &= ~(ST0_CU2|ST0_CU1);
 162
 163 #ifdef CONFIG_MIPS_MT_SMTC
 164         /*
 165          * SMTC restores TCStatus after Status, and the CU bits
 166          * are aliased there.
 167          */
 168         childregs->cp0_tcstatus &= ~(ST0_CU2|ST0_CU1);
 169 #endif
 170         clear_tsk_thread_flag(p, TIF_USEDFPU);
 171
 172 #ifdef CONFIG_MIPS_MT_FPAFF
 173         clear_tsk_thread_flag(p, TIF_FPUBOUND);
 174 #endif /* CONFIG_MIPS_MT_FPAFF */
 175
 176         if (clone_flags & CLONE_SETTLS)
 177                 ti->tp_value = regs->regs[7];
 178
 179         return 0;
 180 }
 181
 182 /* Fill in the fpu structure for a core dump.. */
 183 int dump_fpu(struct pt_regs *regs, elf_fpregset_t *r)
 184 {
 185         memcpy(r, &current->thread.fpu, sizeof(current->thread.fpu));
 186
 187         return 1;
 188 }
 189
 190 void elf_dump_regs(elf_greg_t *gp, struct pt_regs *regs)
 191 {
 192         int i;
 193
 194         for (i = 0; i < EF_R0; i++)
 195                 gp[i] = 0;
 196         gp[EF_R0] = 0;
 197         for (i = 1; i <= 31; i++)
 198                 gp[EF_R0 + i] = regs->regs[i];
 199         gp[EF_R26] = 0;
 200         gp[EF_R27] = 0;
 201         gp[EF_LO] = regs->lo;
 202         gp[EF_HI] = regs->hi;
 203         gp[EF_CP0_EPC] = regs->cp0_epc;
 204         gp[EF_CP0_BADVADDR] = regs->cp0_badvaddr;
 205         gp[EF_CP0_STATUS] = regs->cp0_status;
 206         gp[EF_CP0_CAUSE] = regs->cp0_cause;
 207 #ifdef EF_UNUSED0
 208         gp[EF_UNUSED0] = 0;
 209 #endif
 210 }
 211
 212 int dump_task_regs(struct task_struct *tsk, elf_gregset_t *regs)
 213 {
 214         elf_dump_regs(*regs, task_pt_regs(tsk));
 215         return 1;
 216 }
 217
 218 int dump_task_fpu(struct task_struct *t, elf_fpregset_t *fpr)
 219 {
 220         memcpy(fpr, &t->thread.fpu, sizeof(current->thread.fpu));
 221
 222         return 1;
 223 }
 224
 225 /*
 226  * Create a kernel thread
 227  */
 228 static void __noreturn kernel_thread_helper(void *arg, int (*fn)(void *))
 229 {
 230         do_exit(fn(arg));
 231 }
 232
 233 long kernel_thread(int (*fn)(void *), void *arg, unsigned long flags)
 234 {
 235         struct pt_regs regs;
 236
 237         memset(&regs, 0, sizeof(regs));
 238
 239         regs.regs[4] = (unsigned long) arg;
 240         regs.regs[5] = (unsigned long) fn;
 241         regs.cp0_epc = (unsigned long) kernel_thread_helper;
 242         regs.cp0_status = read_c0_status();
 243 #if defined(CONFIG_CPU_R3000) || defined(CONFIG_CPU_TX39XX)
 244         regs.cp0_status = (regs.cp0_status & ~(ST0_KUP | ST0_IEP | ST0_IEC)) |
 245                           ((regs.cp0_status & (ST0_KUC | ST0_IEC)) << 2);
 246 #else
 247         regs.cp0_status |= ST0_EXL;
 248 #endif
 249
 250         /* Ok, create the new process.. */
 251         return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, &regs, 0, NULL, NULL);
 252 }
 253
 254 /*
 255  *
 256  */
 257 struct mips_frame_info {
 258         void            *func;
 259         unsigned long   func_size;
 260         int             frame_size;
 261         int             pc_offset;
 262 };
 263
 264 static inline int is_ra_save_ins(union mips_instruction *ip)
 265 {
 266         /* sw / sd $ra, offset($sp) */
 267         return (ip->i_format.opcode == sw_op || ip->i_format.opcode == sd_op) &&
 268                 ip->i_format.rs == 29 &&
 269                 ip->i_format.rt == 31;
 270 }
 271
 272 static inline int is_jal_jalr_jr_ins(union mips_instruction *ip)
 273 {
 274         if (ip->j_format.opcode == jal_op)
 275                 return 1;
 276         if (ip->r_format.opcode != spec_op)
 277                 return 0;
 278         return ip->r_format.func == jalr_op || ip->r_format.func == jr_op;
 279 }
 280
 281 static inline int is_sp_move_ins(union mips_instruction *ip)
 282 {
 283         /* addiu/daddiu sp,sp,-imm */
 284         if (ip->i_format.rs != 29 || ip->i_format.rt != 29)
 285                 return 0;
 286         if (ip->i_format.opcode == addiu_op || ip->i_format.opcode == daddiu_op)
 287                 return 1;
 288         return 0;
 289 }
 290
 291 static int get_frame_info(struct mips_frame_info *info)
 292 {
 293         union mips_instruction *ip = info->func;
 294         unsigned max_insns = info->func_size / sizeof(union mips_instruction);
 295         unsigned i;
 296
 297         info->pc_offset = -1;
 298         info->frame_size = 0;
 299
 300         if (!ip)
 301                 goto err;
 302
 303         if (max_insns == 0)
 304                 max_insns = 128U;       /* unknown function size */
 305         max_insns = min(128U, max_insns);
 306
 307         for (i = 0; i < max_insns; i++, ip++) {
 308
 309                 if (is_jal_jalr_jr_ins(ip))
 310                         break;
 311                 if (!info->frame_size) {
 312                         if (is_sp_move_ins(ip))
 313                                 info->frame_size = - ip->i_format.simmediate;
 314                         continue;
 315                 }
 316                 if (info->pc_offset == -1 && is_ra_save_ins(ip)) {
 317                         info->pc_offset =
 318                                 ip->i_format.simmediate / sizeof(long);
 319                         break;
 320                 }
 321         }
 322         if (info->frame_size && info->pc_offset >= 0) /* nested */
 323                 return 0;
 324         if (info->pc_offset < 0) /* leaf */
 325                 return 1;
 326         /* prologue seems boggus... */
 327 err:
 328         return -1;
 329 }
 330
 331 static struct mips_frame_info schedule_mfi __read_mostly;
 332
 333 static int __init frame_info_init(void)
 334 {
 335         unsigned long size = 0;
 336 #ifdef CONFIG_KALLSYMS
 337         unsigned long ofs;
 338
 339         kallsyms_lookup_size_offset((unsigned long)schedule, &size, &ofs);
 340 #endif
 341         schedule_mfi.func = schedule;
 342         schedule_mfi.func_size = size;
 343
 344         get_frame_info(&schedule_mfi);
 345
 346         /*
 347          * Without schedule() frame info, result given by
 348          * thread_saved_pc() and get_wchan() are not reliable.
 349          */
 350         if (schedule_mfi.pc_offset < 0)
 351                 printk("Can't analyze schedule() prologue at %p\n", schedule);
 352
 353         return 0;
 354 }
 355
 356 arch_initcall(frame_info_init);
 357
 358 /*
 359  * Return saved PC of a blocked thread.
 360  */
 361 unsigned long thread_saved_pc(struct task_struct *tsk)
 362 {
 363         struct thread_struct *t = &tsk->thread;
 364
 365         /* New born processes are a special case */
 366         if (t->reg31 == (unsigned long) ret_from_fork)
 367                 return t->reg31;
 368         if (schedule_mfi.pc_offset < 0)
 369                 return 0;
 370         return ((unsigned long *)t->reg29)[schedule_mfi.pc_offset];
 371 }
 372
 373
 374 #ifdef CONFIG_KALLSYMS
 375 /* generic stack unwinding function */
 376 unsigned long notrace unwind_stack_by_address(unsigned long stack_page,
 377                                               unsigned long *sp,
 378                                               unsigned long pc,
 379                                               unsigned long *ra)
 380 {
 381         struct mips_frame_info info;
 382         unsigned long size, ofs;
 383         int leaf;
 384         extern void ret_from_irq(void);
 385         extern void ret_from_exception(void);
 386
 387         if (!stack_page)
 388                 return 0;
 389
 390         /*
 391          * If we reached the bottom of interrupt context,
 392          * return saved pc in pt_regs.
 393          */
 394         if (pc == (unsigned long)ret_from_irq ||
 395             pc == (unsigned long)ret_from_exception) {
 396                 struct pt_regs *regs;
 397                 if (*sp >= stack_page &&
 398                     *sp + sizeof(*regs) <= stack_page + THREAD_SIZE - 32) {
 399                         regs = (struct pt_regs *)*sp;
 400                         pc = regs->cp0_epc;
 401                         if (__kernel_text_address(pc)) {
 402                                 *sp = regs->regs[29];
 403                                 *ra = regs->regs[31];
 404                                 return pc;
 405                         }
 406                 }
 407                 return 0;
 408         }
 409         if (!kallsyms_lookup_size_offset(pc, &size, &ofs))
 410                 return 0;
 411         /*
 412          * Return ra if an exception occurred at the first instruction
 413          */
 414         if (unlikely(ofs == 0)) {
 415                 pc = *ra;
 416                 *ra = 0;
 417                 return pc;
 418         }
 419
 420         info.func = (void *)(pc - ofs);
 421         info.func_size = ofs;   /* analyze from start to ofs */
 422         leaf = get_frame_info(&info);
 423         if (leaf < 0)
 424                 return 0;
 425
 426         if (*sp < stack_page ||
 427             *sp + info.frame_size > stack_page + THREAD_SIZE - 32)
 428                 return 0;
 429
 430         if (leaf)
 431                 /*
 432                  * For some extreme cases, get_frame_info() can
 433                  * consider wrongly a nested function as a leaf
 434                  * one. In that cases avoid to return always the
 435                  * same value.
 436                  */
 437                 pc = pc != *ra ? *ra : 0;
 438         else
 439                 pc = ((unsigned long *)(*sp))[info.pc_offset];
 440
 441         *sp += info.frame_size;
 442         *ra = 0;
 443         return __kernel_text_address(pc) ? pc : 0;
 444 }
 445 EXPORT_SYMBOL(unwind_stack_by_address);
 446
 447 /* used by show_backtrace() */
 448 unsigned long unwind_stack(struct task_struct *task, unsigned long *sp,
 449                            unsigned long pc, unsigned long *ra)
 450 {
 451         unsigned long stack_page = (unsigned long)task_stack_page(task);
 452         return unwind_stack_by_address(stack_page, sp, pc, ra);
 453 }
 454 #endif
 455
 456 /*
 457  * get_wchan - a maintenance nightmare^W^Wpain in the ass ...
 458  */
 459 unsigned long get_wchan(struct task_struct *task)
 460 {
 461         unsigned long pc = 0;
 462 #ifdef CONFIG_KALLSYMS
 463         unsigned long sp;
 464         unsigned long ra = 0;
 465 #endif
 466
 467         if (!task || task == current || task->state == TASK_RUNNING)
 468                 goto out;
 469         if (!task_stack_page(task))
 470                 goto out;
 471
 472         pc = thread_saved_pc(task);
 473
 474 #ifdef CONFIG_KALLSYMS
 475         sp = task->thread.reg29 + schedule_mfi.frame_size;
 476
 477         while (in_sched_functions(pc))
 478                 pc = unwind_stack(task, &sp, pc, &ra);
 479 #endif
 480
 481 out:
 482         return pc;
 483 }
 484
 485 /*
 486  * Don't forget that the stack pointer must be aligned on a 8 bytes
 487  * boundary for 32-bits ABI and 16 bytes for 64-bits ABI.
 488  */
 489 unsigned long arch_align_stack(unsigned long sp)
 490 {
 491         if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space)
 492                 sp -= get_random_int() & ~PAGE_MASK;
 493
 494         return sp & ALMASK;
 495 }