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/module.h>
  13 #include <linux/sched.h>
  14 #include <linux/tick.h>
  15 #include <linux/kernel.h>
  16 #include <linux/mm.h>
  17 #include <linux/stddef.h>
  18 #include <linux/unistd.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_stop_sched_tick(1);
  60                 while (!need_resched() && cpu_online(cpu)) {
  61 #ifdef CONFIG_MIPS_MT_SMTC
  62                         extern void smtc_idle_loop_hook(void);
  63
  64                         smtc_idle_loop_hook();
  65 #endif
  66
  67                         if (cpu_wait) {
  68                                 /* Don't trace irqs off for idle */
  69                                 stop_critical_timings();
  70                                 (*cpu_wait)();
  71                                 start_critical_timings();
  72                         }
  73                 }
  74 #ifdef CONFIG_HOTPLUG_CPU
  75                 if (!cpu_online(cpu) && !cpu_isset(cpu, cpu_callin_map) &&
  76                     (system_state == SYSTEM_RUNNING ||
  77                      system_state == SYSTEM_BOOTING))
  78                         play_dead();
  79 #endif
  80                 tick_nohz_restart_sched_tick();
  81                 preempt_enable_no_resched();
  82                 schedule();
  83                 preempt_disable();
  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         current_thread_info()->addr_limit = USER_DS;
 107 }
 108
 109 void exit_thread(void)
 110 {
 111 }
 112
 113 void flush_thread(void)
 114 {
 115 }
 116
 117 int copy_thread(unsigned long clone_flags, unsigned long usp,
 118         unsigned long unused, struct task_struct *p, struct pt_regs *regs)
 119 {
 120         struct thread_info *ti = task_thread_info(p);
 121         struct pt_regs *childregs;
 122         unsigned long childksp;
 123         p->set_child_tid = p->clear_child_tid = NULL;
 124
 125         childksp = (unsigned long)task_stack_page(p) + THREAD_SIZE - 32;
 126
 127         preempt_disable();
 128
 129         if (is_fpu_owner())
 130                 save_fp(p);
 131
 132         if (cpu_has_dsp)
 133                 save_dsp(p);
 134
 135         preempt_enable();
 136
 137         /* set up new TSS. */
 138         childregs = (struct pt_regs *) childksp - 1;
 139         /*  Put the stack after the struct pt_regs.  */
 140         childksp = (unsigned long) childregs;
 141         *childregs = *regs;
 142         childregs->regs[7] = 0; /* Clear error flag */
 143
 144         childregs->regs[2] = 0; /* Child gets zero as return value */
 145
 146         if (childregs->cp0_status & ST0_CU0) {
 147                 childregs->regs[28] = (unsigned long) ti;
 148                 childregs->regs[29] = childksp;
 149                 ti->addr_limit = KERNEL_DS;
 150         } else {
 151                 childregs->regs[29] = usp;
 152                 ti->addr_limit = USER_DS;
 153         }
 154         p->thread.reg29 = (unsigned long) childregs;
 155         p->thread.reg31 = (unsigned long) ret_from_fork;
 156
 157         /*
 158          * New tasks lose permission to use the fpu. This accelerates context
 159          * switching for most programs since they don't use the fpu.
 160          */
 161         p->thread.cp0_status = read_c0_status() & ~(ST0_CU2|ST0_CU1);
 162         childregs->cp0_status &= ~(ST0_CU2|ST0_CU1);
 163
 164 #ifdef CONFIG_MIPS_MT_SMTC
 165         /*
 166          * SMTC restores TCStatus after Status, and the CU bits
 167          * are aliased there.
 168          */
 169         childregs->cp0_tcstatus &= ~(ST0_CU2|ST0_CU1);
 170 #endif
 171         clear_tsk_thread_flag(p, TIF_USEDFPU);
 172
 173 #ifdef CONFIG_MIPS_MT_FPAFF
 174         clear_tsk_thread_flag(p, TIF_FPUBOUND);
 175 #endif /* CONFIG_MIPS_MT_FPAFF */
 176
 177         if (clone_flags & CLONE_SETTLS)
 178                 ti->tp_value = regs->regs[7];
 179
 180         return 0;
 181 }
 182
 183 /* Fill in the fpu structure for a core dump.. */
 184 int dump_fpu(struct pt_regs *regs, elf_fpregset_t *r)
 185 {
 186         memcpy(r, &current->thread.fpu, sizeof(current->thread.fpu));
 187
 188         return 1;
 189 }
 190
 191 void elf_dump_regs(elf_greg_t *gp, struct pt_regs *regs)
 192 {
 193         int i;
 194
 195         for (i = 0; i < EF_R0; i++)
 196                 gp[i] = 0;
 197         gp[EF_R0] = 0;
 198         for (i = 1; i <= 31; i++)
 199                 gp[EF_R0 + i] = regs->regs[i];
 200         gp[EF_R26] = 0;
 201         gp[EF_R27] = 0;
 202         gp[EF_LO] = regs->lo;
 203         gp[EF_HI] = regs->hi;
 204         gp[EF_CP0_EPC] = regs->cp0_epc;
 205         gp[EF_CP0_BADVADDR] = regs->cp0_badvaddr;
 206         gp[EF_CP0_STATUS] = regs->cp0_status;
 207         gp[EF_CP0_CAUSE] = regs->cp0_cause;
 208 #ifdef EF_UNUSED0
 209         gp[EF_UNUSED0] = 0;
 210 #endif
 211 }
 212
 213 int dump_task_regs(struct task_struct *tsk, elf_gregset_t *regs)
 214 {
 215         elf_dump_regs(*regs, task_pt_regs(tsk));
 216         return 1;
 217 }
 218
 219 int dump_task_fpu(struct task_struct *t, elf_fpregset_t *fpr)
 220 {
 221         memcpy(fpr, &t->thread.fpu, sizeof(current->thread.fpu));
 222
 223         return 1;
 224 }
 225
 226 /*
 227  * Create a kernel thread
 228  */
 229 static void __noreturn kernel_thread_helper(void *arg, int (*fn)(void *))
 230 {
 231         do_exit(fn(arg));
 232 }
 233
 234 long kernel_thread(int (*fn)(void *), void *arg, unsigned long flags)
 235 {
 236         struct pt_regs regs;
 237
 238         memset(&regs, 0, sizeof(regs));
 239
 240         regs.regs[4] = (unsigned long) arg;
 241         regs.regs[5] = (unsigned long) fn;
 242         regs.cp0_epc = (unsigned long) kernel_thread_helper;
 243         regs.cp0_status = read_c0_status();
 244 #if defined(CONFIG_CPU_R3000) || defined(CONFIG_CPU_TX39XX)
 245         regs.cp0_status = (regs.cp0_status & ~(ST0_KUP | ST0_IEP | ST0_IEC)) |
 246                           ((regs.cp0_status & (ST0_KUC | ST0_IEC)) << 2);
 247 #else
 248         regs.cp0_status |= ST0_EXL;
 249 #endif
 250
 251         /* Ok, create the new process.. */
 252         return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, &regs, 0, NULL, NULL);
 253 }
 254
 255 /*
 256  *
 257  */
 258 struct mips_frame_info {
 259         void            *func;
 260         unsigned long   func_size;
 261         int             frame_size;
 262         int             pc_offset;
 263 };
 264
 265 static inline int is_ra_save_ins(union mips_instruction *ip)
 266 {
 267         /* sw / sd $ra, offset($sp) */
 268         return (ip->i_format.opcode == sw_op || ip->i_format.opcode == sd_op) &&
 269                 ip->i_format.rs == 29 &&
 270                 ip->i_format.rt == 31;
 271 }
 272
 273 static inline int is_jal_jalr_jr_ins(union mips_instruction *ip)
 274 {
 275         if (ip->j_format.opcode == jal_op)
 276                 return 1;
 277         if (ip->r_format.opcode != spec_op)
 278                 return 0;
 279         return ip->r_format.func == jalr_op || ip->r_format.func == jr_op;
 280 }
 281
 282 static inline int is_sp_move_ins(union mips_instruction *ip)
 283 {
 284         /* addiu/daddiu sp,sp,-imm */
 285         if (ip->i_format.rs != 29 || ip->i_format.rt != 29)
 286                 return 0;
 287         if (ip->i_format.opcode == addiu_op || ip->i_format.opcode == daddiu_op)
 288                 return 1;
 289         return 0;
 290 }
 291
 292 static int get_frame_info(struct mips_frame_info *info)
 293 {
 294         union mips_instruction *ip = info->func;
 295         unsigned max_insns = info->func_size / sizeof(union mips_instruction);
 296         unsigned i;
 297
 298         info->pc_offset = -1;
 299         info->frame_size = 0;
 300
 301         if (!ip)
 302                 goto err;
 303
 304         if (max_insns == 0)
 305                 max_insns = 128U;       /* unknown function size */
 306         max_insns = min(128U, max_insns);
 307
 308         for (i = 0; i < max_insns; i++, ip++) {
 309
 310                 if (is_jal_jalr_jr_ins(ip))
 311                         break;
 312                 if (!info->frame_size) {
 313                         if (is_sp_move_ins(ip))
 314                                 info->frame_size = - ip->i_format.simmediate;
 315                         continue;
 316                 }
 317                 if (info->pc_offset == -1 && is_ra_save_ins(ip)) {
 318                         info->pc_offset =
 319                                 ip->i_format.simmediate / sizeof(long);
 320                         break;
 321                 }
 322         }
 323         if (info->frame_size && info->pc_offset >= 0) /* nested */
 324                 return 0;
 325         if (info->pc_offset < 0) /* leaf */
 326                 return 1;
 327         /* prologue seems boggus... */
 328 err:
 329         return -1;
 330 }
 331
 332 static struct mips_frame_info schedule_mfi __read_mostly;
 333
 334 static int __init frame_info_init(void)
 335 {
 336         unsigned long size = 0;
 337 #ifdef CONFIG_KALLSYMS
 338         unsigned long ofs;
 339
 340         kallsyms_lookup_size_offset((unsigned long)schedule, &size, &ofs);
 341 #endif
 342         schedule_mfi.func = schedule;
 343         schedule_mfi.func_size = size;
 344
 345         get_frame_info(&schedule_mfi);
 346
 347         /*
 348          * Without schedule() frame info, result given by
 349          * thread_saved_pc() and get_wchan() are not reliable.
 350          */
 351         if (schedule_mfi.pc_offset < 0)
 352                 printk("Can't analyze schedule() prologue at %p\n", schedule);
 353
 354         return 0;
 355 }
 356
 357 arch_initcall(frame_info_init);
 358
 359 /*
 360  * Return saved PC of a blocked thread.
 361  */
 362 unsigned long thread_saved_pc(struct task_struct *tsk)
 363 {
 364         struct thread_struct *t = &tsk->thread;
 365
 366         /* New born processes are a special case */
 367         if (t->reg31 == (unsigned long) ret_from_fork)
 368                 return t->reg31;
 369         if (schedule_mfi.pc_offset < 0)
 370                 return 0;
 371         return ((unsigned long *)t->reg29)[schedule_mfi.pc_offset];
 372 }
 373
 374
 375 #ifdef CONFIG_KALLSYMS
 376 /* used by show_backtrace() */
 377 unsigned long unwind_stack(struct task_struct *task, unsigned long *sp,
 378                            unsigned long pc, unsigned long *ra)
 379 {
 380         unsigned long stack_page;
 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         stack_page = (unsigned long)task_stack_page(task);
 388         if (!stack_page)
 389                 return 0;
 390
 391         /*
 392          * If we reached the bottom of interrupt context,
 393          * return saved pc in pt_regs.
 394          */
 395         if (pc == (unsigned long)ret_from_irq ||
 396             pc == (unsigned long)ret_from_exception) {
 397                 struct pt_regs *regs;
 398                 if (*sp >= stack_page &&
 399                     *sp + sizeof(*regs) <= stack_page + THREAD_SIZE - 32) {
 400                         regs = (struct pt_regs *)*sp;
 401                         pc = regs->cp0_epc;
 402                         if (__kernel_text_address(pc)) {
 403                                 *sp = regs->regs[29];
 404                                 *ra = regs->regs[31];
 405                                 return pc;
 406                         }
 407                 }
 408                 return 0;
 409         }
 410         if (!kallsyms_lookup_size_offset(pc, &size, &ofs))
 411                 return 0;
 412         /*
 413          * Return ra if an exception occurred at the first instruction
 414          */
 415         if (unlikely(ofs == 0)) {
 416                 pc = *ra;
 417                 *ra = 0;
 418                 return pc;
 419         }
 420
 421         info.func = (void *)(pc - ofs);
 422         info.func_size = ofs;   /* analyze from start to ofs */
 423         leaf = get_frame_info(&info);
 424         if (leaf < 0)
 425                 return 0;
 426
 427         if (*sp < stack_page ||
 428             *sp + info.frame_size > stack_page + THREAD_SIZE - 32)
 429                 return 0;
 430
 431         if (leaf)
 432                 /*
 433                  * For some extreme cases, get_frame_info() can
 434                  * consider wrongly a nested function as a leaf
 435                  * one. In that cases avoid to return always the
 436                  * same value.
 437                  */
 438                 pc = pc != *ra ? *ra : 0;
 439         else
 440                 pc = ((unsigned long *)(*sp))[info.pc_offset];
 441
 442         *sp += info.frame_size;
 443         *ra = 0;
 444         return __kernel_text_address(pc) ? pc : 0;
 445 }
 446 #endif
 447
 448 /*
 449  * get_wchan - a maintenance nightmare^W^Wpain in the ass ...
 450  */
 451 unsigned long get_wchan(struct task_struct *task)
 452 {
 453         unsigned long pc = 0;
 454 #ifdef CONFIG_KALLSYMS
 455         unsigned long sp;
 456         unsigned long ra = 0;
 457 #endif
 458
 459         if (!task || task == current || task->state == TASK_RUNNING)
 460                 goto out;
 461         if (!task_stack_page(task))
 462                 goto out;
 463
 464         pc = thread_saved_pc(task);
 465
 466 #ifdef CONFIG_KALLSYMS
 467         sp = task->thread.reg29 + schedule_mfi.frame_size;
 468
 469         while (in_sched_functions(pc))
 470                 pc = unwind_stack(task, &sp, pc, &ra);
 471 #endif
 472
 473 out:
 474         return pc;
 475 }
 476
 477 /*
 478  * Don't forget that the stack pointer must be aligned on a 8 bytes
 479  * boundary for 32-bits ABI and 16 bytes for 64-bits ABI.
 480  */
 481 unsigned long arch_align_stack(unsigned long sp)
 482 {
 483         if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space)
 484                 sp -= get_random_int() & ~PAGE_MASK;
 485
 486         return sp & ALMASK;
 487 }