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