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