1 /* 2 * Architecture-specific setup. 3 * 4 * Copyright (C) 1998-2003 Hewlett-Packard Co 5 * David Mosberger-Tang <davidm@hpl.hp.com> 6 * 04/11/17 Ashok Raj <ashok.raj@intel.com> Added CPU Hotplug Support 7 * 8 * 2005-10-07 Keith Owens <kaos@sgi.com> 9 * Add notify_die() hooks. 10 */ 11 #include <linux/cpu.h> 12 #include <linux/pm.h> 13 #include <linux/elf.h> 14 #include <linux/errno.h> 15 #include <linux/kallsyms.h> 16 #include <linux/kernel.h> 17 #include <linux/mm.h> 18 #include <linux/module.h> 19 #include <linux/notifier.h> 20 #include <linux/personality.h> 21 #include <linux/sched.h> 22 #include <linux/slab.h> 23 #include <linux/stddef.h> 24 #include <linux/thread_info.h> 25 #include <linux/unistd.h> 26 #include <linux/efi.h> 27 #include <linux/interrupt.h> 28 #include <linux/delay.h> 29 #include <linux/kdebug.h> 30 #include <linux/utsname.h> 31 #include <linux/tracehook.h> 32 33 #include <asm/cpu.h> 34 #include <asm/delay.h> 35 #include <asm/elf.h> 36 #include <asm/ia32.h> 37 #include <asm/irq.h> 38 #include <asm/kexec.h> 39 #include <asm/pgalloc.h> 40 #include <asm/processor.h> 41 #include <asm/sal.h> 42 #include <asm/tlbflush.h> 43 #include <asm/uaccess.h> 44 #include <asm/unwind.h> 45 #include <asm/user.h> 46 47 #include "entry.h" 48 49 #ifdef CONFIG_PERFMON 50 # include <asm/perfmon.h> 51 #endif 52 53 #include "sigframe.h" 54 55 void (*ia64_mark_idle)(int); 56 57 unsigned long boot_option_idle_override = 0; 58 EXPORT_SYMBOL(boot_option_idle_override); 59 unsigned long idle_halt; 60 EXPORT_SYMBOL(idle_halt); 61 unsigned long idle_nomwait; 62 EXPORT_SYMBOL(idle_nomwait); 63 64 void 65 ia64_do_show_stack (struct unw_frame_info *info, void *arg) 66 { 67 unsigned long ip, sp, bsp; 68 char buf[128]; /* don't make it so big that it overflows the stack! */ 69 70 printk("\nCall Trace:\n"); 71 do { 72 unw_get_ip(info, &ip); 73 if (ip == 0) 74 break; 75 76 unw_get_sp(info, &sp); 77 unw_get_bsp(info, &bsp); 78 snprintf(buf, sizeof(buf), 79 " [<%016lx>] %%s\n" 80 " sp=%016lx bsp=%016lx\n", 81 ip, sp, bsp); 82 print_symbol(buf, ip); 83 } while (unw_unwind(info) >= 0); 84 } 85 86 void 87 show_stack (struct task_struct *task, unsigned long *sp) 88 { 89 if (!task) 90 unw_init_running(ia64_do_show_stack, NULL); 91 else { 92 struct unw_frame_info info; 93 94 unw_init_from_blocked_task(&info, task); 95 ia64_do_show_stack(&info, NULL); 96 } 97 } 98 99 void 100 dump_stack (void) 101 { 102 show_stack(NULL, NULL); 103 } 104 105 EXPORT_SYMBOL(dump_stack); 106 107 void 108 show_regs (struct pt_regs *regs) 109 { 110 unsigned long ip = regs->cr_iip + ia64_psr(regs)->ri; 111 112 print_modules(); 113 printk("\nPid: %d, CPU %d, comm: %20s\n", task_pid_nr(current), 114 smp_processor_id(), current->comm); 115 printk("psr : %016lx ifs : %016lx ip : [<%016lx>] %s (%s)\n", 116 regs->cr_ipsr, regs->cr_ifs, ip, print_tainted(), 117 init_utsname()->release); 118 print_symbol("ip is at %s\n", ip); 119 printk("unat: %016lx pfs : %016lx rsc : %016lx\n", 120 regs->ar_unat, regs->ar_pfs, regs->ar_rsc); 121 printk("rnat: %016lx bsps: %016lx pr : %016lx\n", 122 regs->ar_rnat, regs->ar_bspstore, regs->pr); 123 printk("ldrs: %016lx ccv : %016lx fpsr: %016lx\n", 124 regs->loadrs, regs->ar_ccv, regs->ar_fpsr); 125 printk("csd : %016lx ssd : %016lx\n", regs->ar_csd, regs->ar_ssd); 126 printk("b0 : %016lx b6 : %016lx b7 : %016lx\n", regs->b0, regs->b6, regs->b7); 127 printk("f6 : %05lx%016lx f7 : %05lx%016lx\n", 128 regs->f6.u.bits[1], regs->f6.u.bits[0], 129 regs->f7.u.bits[1], regs->f7.u.bits[0]); 130 printk("f8 : %05lx%016lx f9 : %05lx%016lx\n", 131 regs->f8.u.bits[1], regs->f8.u.bits[0], 132 regs->f9.u.bits[1], regs->f9.u.bits[0]); 133 printk("f10 : %05lx%016lx f11 : %05lx%016lx\n", 134 regs->f10.u.bits[1], regs->f10.u.bits[0], 135 regs->f11.u.bits[1], regs->f11.u.bits[0]); 136 137 printk("r1 : %016lx r2 : %016lx r3 : %016lx\n", regs->r1, regs->r2, regs->r3); 138 printk("r8 : %016lx r9 : %016lx r10 : %016lx\n", regs->r8, regs->r9, regs->r10); 139 printk("r11 : %016lx r12 : %016lx r13 : %016lx\n", regs->r11, regs->r12, regs->r13); 140 printk("r14 : %016lx r15 : %016lx r16 : %016lx\n", regs->r14, regs->r15, regs->r16); 141 printk("r17 : %016lx r18 : %016lx r19 : %016lx\n", regs->r17, regs->r18, regs->r19); 142 printk("r20 : %016lx r21 : %016lx r22 : %016lx\n", regs->r20, regs->r21, regs->r22); 143 printk("r23 : %016lx r24 : %016lx r25 : %016lx\n", regs->r23, regs->r24, regs->r25); 144 printk("r26 : %016lx r27 : %016lx r28 : %016lx\n", regs->r26, regs->r27, regs->r28); 145 printk("r29 : %016lx r30 : %016lx r31 : %016lx\n", regs->r29, regs->r30, regs->r31); 146 147 if (user_mode(regs)) { 148 /* print the stacked registers */ 149 unsigned long val, *bsp, ndirty; 150 int i, sof, is_nat = 0; 151 152 sof = regs->cr_ifs & 0x7f; /* size of frame */ 153 ndirty = (regs->loadrs >> 19); 154 bsp = ia64_rse_skip_regs((unsigned long *) regs->ar_bspstore, ndirty); 155 for (i = 0; i < sof; ++i) { 156 get_user(val, (unsigned long __user *) ia64_rse_skip_regs(bsp, i)); 157 printk("r%-3u:%c%016lx%s", 32 + i, is_nat ? '*' : ' ', val, 158 ((i == sof - 1) || (i % 3) == 2) ? "\n" : " "); 159 } 160 } else 161 show_stack(NULL, NULL); 162 } 163 164 void 165 do_notify_resume_user(sigset_t *unused, struct sigscratch *scr, long in_syscall) 166 { 167 if (fsys_mode(current, &scr->pt)) { 168 /* 169 * defer signal-handling etc. until we return to 170 * privilege-level 0. 171 */ 172 if (!ia64_psr(&scr->pt)->lp) 173 ia64_psr(&scr->pt)->lp = 1; 174 return; 175 } 176 177 #ifdef CONFIG_PERFMON 178 if (current->thread.pfm_needs_checking) 179 /* 180 * Note: pfm_handle_work() allow us to call it with interrupts 181 * disabled, and may enable interrupts within the function. 182 */ 183 pfm_handle_work(); 184 #endif 185 186 /* deal with pending signal delivery */ 187 if (test_thread_flag(TIF_SIGPENDING)) { 188 local_irq_enable(); /* force interrupt enable */ 189 ia64_do_signal(scr, in_syscall); 190 } 191 192 if (test_thread_flag(TIF_NOTIFY_RESUME)) { 193 clear_thread_flag(TIF_NOTIFY_RESUME); 194 tracehook_notify_resume(&scr->pt); 195 } 196 197 /* copy user rbs to kernel rbs */ 198 if (unlikely(test_thread_flag(TIF_RESTORE_RSE))) { 199 local_irq_enable(); /* force interrupt enable */ 200 ia64_sync_krbs(); 201 } 202 203 local_irq_disable(); /* force interrupt disable */ 204 } 205 206 static int pal_halt = 1; 207 static int can_do_pal_halt = 1; 208 209 static int __init nohalt_setup(char * str) 210 { 211 pal_halt = can_do_pal_halt = 0; 212 return 1; 213 } 214 __setup("nohalt", nohalt_setup); 215 216 void 217 update_pal_halt_status(int status) 218 { 219 can_do_pal_halt = pal_halt && status; 220 } 221 222 /* 223 * We use this if we don't have any better idle routine.. 224 */ 225 void 226 default_idle (void) 227 { 228 local_irq_enable(); 229 while (!need_resched()) { 230 if (can_do_pal_halt) { 231 local_irq_disable(); 232 if (!need_resched()) { 233 safe_halt(); 234 } 235 local_irq_enable(); 236 } else 237 cpu_relax(); 238 } 239 } 240 241 #ifdef CONFIG_HOTPLUG_CPU 242 /* We don't actually take CPU down, just spin without interrupts. */ 243 static inline void play_dead(void) 244 { 245 unsigned int this_cpu = smp_processor_id(); 246 247 /* Ack it */ 248 __get_cpu_var(cpu_state) = CPU_DEAD; 249 250 max_xtp(); 251 local_irq_disable(); 252 idle_task_exit(); 253 ia64_jump_to_sal(&sal_boot_rendez_state[this_cpu]); 254 /* 255 * The above is a point of no-return, the processor is 256 * expected to be in SAL loop now. 257 */ 258 BUG(); 259 } 260 #else 261 static inline void play_dead(void) 262 { 263 BUG(); 264 } 265 #endif /* CONFIG_HOTPLUG_CPU */ 266 267 static void do_nothing(void *unused) 268 { 269 } 270 271 /* 272 * cpu_idle_wait - Used to ensure that all the CPUs discard old value of 273 * pm_idle and update to new pm_idle value. Required while changing pm_idle 274 * handler on SMP systems. 275 * 276 * Caller must have changed pm_idle to the new value before the call. Old 277 * pm_idle value will not be used by any CPU after the return of this function. 278 */ 279 void cpu_idle_wait(void) 280 { 281 smp_mb(); 282 /* kick all the CPUs so that they exit out of pm_idle */ 283 smp_call_function(do_nothing, NULL, 1); 284 } 285 EXPORT_SYMBOL_GPL(cpu_idle_wait); 286 287 void __attribute__((noreturn)) 288 cpu_idle (void) 289 { 290 void (*mark_idle)(int) = ia64_mark_idle; 291 int cpu = smp_processor_id(); 292 293 /* endless idle loop with no priority at all */ 294 while (1) { 295 if (can_do_pal_halt) { 296 current_thread_info()->status &= ~TS_POLLING; 297 /* 298 * TS_POLLING-cleared state must be visible before we 299 * test NEED_RESCHED: 300 */ 301 smp_mb(); 302 } else { 303 current_thread_info()->status |= TS_POLLING; 304 } 305 306 if (!need_resched()) { 307 void (*idle)(void); 308 #ifdef CONFIG_SMP 309 min_xtp(); 310 #endif 311 rmb(); 312 if (mark_idle) 313 (*mark_idle)(1); 314 315 idle = pm_idle; 316 if (!idle) 317 idle = default_idle; 318 (*idle)(); 319 if (mark_idle) 320 (*mark_idle)(0); 321 #ifdef CONFIG_SMP 322 normal_xtp(); 323 #endif 324 } 325 preempt_enable_no_resched(); 326 schedule(); 327 preempt_disable(); 328 check_pgt_cache(); 329 if (cpu_is_offline(cpu)) 330 play_dead(); 331 } 332 } 333 334 void 335 ia64_save_extra (struct task_struct *task) 336 { 337 #ifdef CONFIG_PERFMON 338 unsigned long info; 339 #endif 340 341 if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0) 342 ia64_save_debug_regs(&task->thread.dbr[0]); 343 344 #ifdef CONFIG_PERFMON 345 if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0) 346 pfm_save_regs(task); 347 348 info = __get_cpu_var(pfm_syst_info); 349 if (info & PFM_CPUINFO_SYST_WIDE) 350 pfm_syst_wide_update_task(task, info, 0); 351 #endif 352 353 #ifdef CONFIG_IA32_SUPPORT 354 if (IS_IA32_PROCESS(task_pt_regs(task))) 355 ia32_save_state(task); 356 #endif 357 } 358 359 void 360 ia64_load_extra (struct task_struct *task) 361 { 362 #ifdef CONFIG_PERFMON 363 unsigned long info; 364 #endif 365 366 if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0) 367 ia64_load_debug_regs(&task->thread.dbr[0]); 368 369 #ifdef CONFIG_PERFMON 370 if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0) 371 pfm_load_regs(task); 372 373 info = __get_cpu_var(pfm_syst_info); 374 if (info & PFM_CPUINFO_SYST_WIDE) 375 pfm_syst_wide_update_task(task, info, 1); 376 #endif 377 378 #ifdef CONFIG_IA32_SUPPORT 379 if (IS_IA32_PROCESS(task_pt_regs(task))) 380 ia32_load_state(task); 381 #endif 382 } 383 384 /* 385 * Copy the state of an ia-64 thread. 386 * 387 * We get here through the following call chain: 388 * 389 * from user-level: from kernel: 390 * 391 * <clone syscall> <some kernel call frames> 392 * sys_clone : 393 * do_fork do_fork 394 * copy_thread copy_thread 395 * 396 * This means that the stack layout is as follows: 397 * 398 * +---------------------+ (highest addr) 399 * | struct pt_regs | 400 * +---------------------+ 401 * | struct switch_stack | 402 * +---------------------+ 403 * | | 404 * | memory stack | 405 * | | <-- sp (lowest addr) 406 * +---------------------+ 407 * 408 * Observe that we copy the unat values that are in pt_regs and switch_stack. Spilling an 409 * integer to address X causes bit N in ar.unat to be set to the NaT bit of the register, 410 * with N=(X & 0x1ff)/8. Thus, copying the unat value preserves the NaT bits ONLY if the 411 * pt_regs structure in the parent is congruent to that of the child, modulo 512. Since 412 * the stack is page aligned and the page size is at least 4KB, this is always the case, 413 * so there is nothing to worry about. 414 */ 415 int 416 copy_thread (int nr, unsigned long clone_flags, 417 unsigned long user_stack_base, unsigned long user_stack_size, 418 struct task_struct *p, struct pt_regs *regs) 419 { 420 extern char ia64_ret_from_clone, ia32_ret_from_clone; 421 struct switch_stack *child_stack, *stack; 422 unsigned long rbs, child_rbs, rbs_size; 423 struct pt_regs *child_ptregs; 424 int retval = 0; 425 426 #ifdef CONFIG_SMP 427 /* 428 * For SMP idle threads, fork_by_hand() calls do_fork with 429 * NULL regs. 430 */ 431 if (!regs) 432 return 0; 433 #endif 434 435 stack = ((struct switch_stack *) regs) - 1; 436 437 child_ptregs = (struct pt_regs *) ((unsigned long) p + IA64_STK_OFFSET) - 1; 438 child_stack = (struct switch_stack *) child_ptregs - 1; 439 440 /* copy parent's switch_stack & pt_regs to child: */ 441 memcpy(child_stack, stack, sizeof(*child_ptregs) + sizeof(*child_stack)); 442 443 rbs = (unsigned long) current + IA64_RBS_OFFSET; 444 child_rbs = (unsigned long) p + IA64_RBS_OFFSET; 445 rbs_size = stack->ar_bspstore - rbs; 446 447 /* copy the parent's register backing store to the child: */ 448 memcpy((void *) child_rbs, (void *) rbs, rbs_size); 449 450 if (likely(user_mode(child_ptregs))) { 451 if ((clone_flags & CLONE_SETTLS) && !IS_IA32_PROCESS(regs)) 452 child_ptregs->r13 = regs->r16; /* see sys_clone2() in entry.S */ 453 if (user_stack_base) { 454 child_ptregs->r12 = user_stack_base + user_stack_size - 16; 455 child_ptregs->ar_bspstore = user_stack_base; 456 child_ptregs->ar_rnat = 0; 457 child_ptregs->loadrs = 0; 458 } 459 } else { 460 /* 461 * Note: we simply preserve the relative position of 462 * the stack pointer here. There is no need to 463 * allocate a scratch area here, since that will have 464 * been taken care of by the caller of sys_clone() 465 * already. 466 */ 467 child_ptregs->r12 = (unsigned long) child_ptregs - 16; /* kernel sp */ 468 child_ptregs->r13 = (unsigned long) p; /* set `current' pointer */ 469 } 470 child_stack->ar_bspstore = child_rbs + rbs_size; 471 if (IS_IA32_PROCESS(regs)) 472 child_stack->b0 = (unsigned long) &ia32_ret_from_clone; 473 else 474 child_stack->b0 = (unsigned long) &ia64_ret_from_clone; 475 476 /* copy parts of thread_struct: */ 477 p->thread.ksp = (unsigned long) child_stack - 16; 478 479 /* stop some PSR bits from being inherited. 480 * the psr.up/psr.pp bits must be cleared on fork but inherited on execve() 481 * therefore we must specify them explicitly here and not include them in 482 * IA64_PSR_BITS_TO_CLEAR. 483 */ 484 child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET) 485 & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP)); 486 487 /* 488 * NOTE: The calling convention considers all floating point 489 * registers in the high partition (fph) to be scratch. Since 490 * the only way to get to this point is through a system call, 491 * we know that the values in fph are all dead. Hence, there 492 * is no need to inherit the fph state from the parent to the 493 * child and all we have to do is to make sure that 494 * IA64_THREAD_FPH_VALID is cleared in the child. 495 * 496 * XXX We could push this optimization a bit further by 497 * clearing IA64_THREAD_FPH_VALID on ANY system call. 498 * However, it's not clear this is worth doing. Also, it 499 * would be a slight deviation from the normal Linux system 500 * call behavior where scratch registers are preserved across 501 * system calls (unless used by the system call itself). 502 */ 503 # define THREAD_FLAGS_TO_CLEAR (IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID \ 504 | IA64_THREAD_PM_VALID) 505 # define THREAD_FLAGS_TO_SET 0 506 p->thread.flags = ((current->thread.flags & ~THREAD_FLAGS_TO_CLEAR) 507 | THREAD_FLAGS_TO_SET); 508 ia64_drop_fpu(p); /* don't pick up stale state from a CPU's fph */ 509 #ifdef CONFIG_IA32_SUPPORT 510 /* 511 * If we're cloning an IA32 task then save the IA32 extra 512 * state from the current task to the new task 513 */ 514 if (IS_IA32_PROCESS(task_pt_regs(current))) { 515 ia32_save_state(p); 516 if (clone_flags & CLONE_SETTLS) 517 retval = ia32_clone_tls(p, child_ptregs); 518 519 /* Copy partially mapped page list */ 520 if (!retval) 521 retval = ia32_copy_ia64_partial_page_list(p, 522 clone_flags); 523 } 524 #endif 525 526 #ifdef CONFIG_PERFMON 527 if (current->thread.pfm_context) 528 pfm_inherit(p, child_ptregs); 529 #endif 530 return retval; 531 } 532 533 static void 534 do_copy_task_regs (struct task_struct *task, struct unw_frame_info *info, void *arg) 535 { 536 unsigned long mask, sp, nat_bits = 0, ar_rnat, urbs_end, cfm; 537 unsigned long uninitialized_var(ip); /* GCC be quiet */ 538 elf_greg_t *dst = arg; 539 struct pt_regs *pt; 540 char nat; 541 int i; 542 543 memset(dst, 0, sizeof(elf_gregset_t)); /* don't leak any kernel bits to user-level */ 544 545 if (unw_unwind_to_user(info) < 0) 546 return; 547 548 unw_get_sp(info, &sp); 549 pt = (struct pt_regs *) (sp + 16); 550 551 urbs_end = ia64_get_user_rbs_end(task, pt, &cfm); 552 553 if (ia64_sync_user_rbs(task, info->sw, pt->ar_bspstore, urbs_end) < 0) 554 return; 555 556 ia64_peek(task, info->sw, urbs_end, (long) ia64_rse_rnat_addr((long *) urbs_end), 557 &ar_rnat); 558 559 /* 560 * coredump format: 561 * r0-r31 562 * NaT bits (for r0-r31; bit N == 1 iff rN is a NaT) 563 * predicate registers (p0-p63) 564 * b0-b7 565 * ip cfm user-mask 566 * ar.rsc ar.bsp ar.bspstore ar.rnat 567 * ar.ccv ar.unat ar.fpsr ar.pfs ar.lc ar.ec 568 */ 569 570 /* r0 is zero */ 571 for (i = 1, mask = (1UL << i); i < 32; ++i) { 572 unw_get_gr(info, i, &dst[i], &nat); 573 if (nat) 574 nat_bits |= mask; 575 mask <<= 1; 576 } 577 dst[32] = nat_bits; 578 unw_get_pr(info, &dst[33]); 579 580 for (i = 0; i < 8; ++i) 581 unw_get_br(info, i, &dst[34 + i]); 582 583 unw_get_rp(info, &ip); 584 dst[42] = ip + ia64_psr(pt)->ri; 585 dst[43] = cfm; 586 dst[44] = pt->cr_ipsr & IA64_PSR_UM; 587 588 unw_get_ar(info, UNW_AR_RSC, &dst[45]); 589 /* 590 * For bsp and bspstore, unw_get_ar() would return the kernel 591 * addresses, but we need the user-level addresses instead: 592 */ 593 dst[46] = urbs_end; /* note: by convention PT_AR_BSP points to the end of the urbs! */ 594 dst[47] = pt->ar_bspstore; 595 dst[48] = ar_rnat; 596 unw_get_ar(info, UNW_AR_CCV, &dst[49]); 597 unw_get_ar(info, UNW_AR_UNAT, &dst[50]); 598 unw_get_ar(info, UNW_AR_FPSR, &dst[51]); 599 dst[52] = pt->ar_pfs; /* UNW_AR_PFS is == to pt->cr_ifs for interrupt frames */ 600 unw_get_ar(info, UNW_AR_LC, &dst[53]); 601 unw_get_ar(info, UNW_AR_EC, &dst[54]); 602 unw_get_ar(info, UNW_AR_CSD, &dst[55]); 603 unw_get_ar(info, UNW_AR_SSD, &dst[56]); 604 } 605 606 void 607 do_dump_task_fpu (struct task_struct *task, struct unw_frame_info *info, void *arg) 608 { 609 elf_fpreg_t *dst = arg; 610 int i; 611 612 memset(dst, 0, sizeof(elf_fpregset_t)); /* don't leak any "random" bits */ 613 614 if (unw_unwind_to_user(info) < 0) 615 return; 616 617 /* f0 is 0.0, f1 is 1.0 */ 618 619 for (i = 2; i < 32; ++i) 620 unw_get_fr(info, i, dst + i); 621 622 ia64_flush_fph(task); 623 if ((task->thread.flags & IA64_THREAD_FPH_VALID) != 0) 624 memcpy(dst + 32, task->thread.fph, 96*16); 625 } 626 627 void 628 do_copy_regs (struct unw_frame_info *info, void *arg) 629 { 630 do_copy_task_regs(current, info, arg); 631 } 632 633 void 634 do_dump_fpu (struct unw_frame_info *info, void *arg) 635 { 636 do_dump_task_fpu(current, info, arg); 637 } 638 639 void 640 ia64_elf_core_copy_regs (struct pt_regs *pt, elf_gregset_t dst) 641 { 642 unw_init_running(do_copy_regs, dst); 643 } 644 645 int 646 dump_fpu (struct pt_regs *pt, elf_fpregset_t dst) 647 { 648 unw_init_running(do_dump_fpu, dst); 649 return 1; /* f0-f31 are always valid so we always return 1 */ 650 } 651 652 long 653 sys_execve (char __user *filename, char __user * __user *argv, char __user * __user *envp, 654 struct pt_regs *regs) 655 { 656 char *fname; 657 int error; 658 659 fname = getname(filename); 660 error = PTR_ERR(fname); 661 if (IS_ERR(fname)) 662 goto out; 663 error = do_execve(fname, argv, envp, regs); 664 putname(fname); 665 out: 666 return error; 667 } 668 669 pid_t 670 kernel_thread (int (*fn)(void *), void *arg, unsigned long flags) 671 { 672 extern void start_kernel_thread (void); 673 unsigned long *helper_fptr = (unsigned long *) &start_kernel_thread; 674 struct { 675 struct switch_stack sw; 676 struct pt_regs pt; 677 } regs; 678 679 memset(®s, 0, sizeof(regs)); 680 regs.pt.cr_iip = helper_fptr[0]; /* set entry point (IP) */ 681 regs.pt.r1 = helper_fptr[1]; /* set GP */ 682 regs.pt.r9 = (unsigned long) fn; /* 1st argument */ 683 regs.pt.r11 = (unsigned long) arg; /* 2nd argument */ 684 /* Preserve PSR bits, except for bits 32-34 and 37-45, which we can't read. */ 685 regs.pt.cr_ipsr = ia64_getreg(_IA64_REG_PSR) | IA64_PSR_BN; 686 regs.pt.cr_ifs = 1UL << 63; /* mark as valid, empty frame */ 687 regs.sw.ar_fpsr = regs.pt.ar_fpsr = ia64_getreg(_IA64_REG_AR_FPSR); 688 regs.sw.ar_bspstore = (unsigned long) current + IA64_RBS_OFFSET; 689 regs.sw.pr = (1 << PRED_KERNEL_STACK); 690 return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, ®s.pt, 0, NULL, NULL); 691 } 692 EXPORT_SYMBOL(kernel_thread); 693 694 /* This gets called from kernel_thread() via ia64_invoke_thread_helper(). */ 695 int 696 kernel_thread_helper (int (*fn)(void *), void *arg) 697 { 698 #ifdef CONFIG_IA32_SUPPORT 699 if (IS_IA32_PROCESS(task_pt_regs(current))) { 700 /* A kernel thread is always a 64-bit process. */ 701 current->thread.map_base = DEFAULT_MAP_BASE; 702 current->thread.task_size = DEFAULT_TASK_SIZE; 703 ia64_set_kr(IA64_KR_IO_BASE, current->thread.old_iob); 704 ia64_set_kr(IA64_KR_TSSD, current->thread.old_k1); 705 } 706 #endif 707 return (*fn)(arg); 708 } 709 710 /* 711 * Flush thread state. This is called when a thread does an execve(). 712 */ 713 void 714 flush_thread (void) 715 { 716 /* drop floating-point and debug-register state if it exists: */ 717 current->thread.flags &= ~(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID); 718 ia64_drop_fpu(current); 719 #ifdef CONFIG_IA32_SUPPORT 720 if (IS_IA32_PROCESS(task_pt_regs(current))) { 721 ia32_drop_ia64_partial_page_list(current); 722 current->thread.task_size = IA32_PAGE_OFFSET; 723 set_fs(USER_DS); 724 memset(current->thread.tls_array, 0, sizeof(current->thread.tls_array)); 725 } 726 #endif 727 } 728 729 /* 730 * Clean up state associated with current thread. This is called when 731 * the thread calls exit(). 732 */ 733 void 734 exit_thread (void) 735 { 736 737 ia64_drop_fpu(current); 738 #ifdef CONFIG_PERFMON 739 /* if needed, stop monitoring and flush state to perfmon context */ 740 if (current->thread.pfm_context) 741 pfm_exit_thread(current); 742 743 /* free debug register resources */ 744 if (current->thread.flags & IA64_THREAD_DBG_VALID) 745 pfm_release_debug_registers(current); 746 #endif 747 if (IS_IA32_PROCESS(task_pt_regs(current))) 748 ia32_drop_ia64_partial_page_list(current); 749 } 750 751 unsigned long 752 get_wchan (struct task_struct *p) 753 { 754 struct unw_frame_info info; 755 unsigned long ip; 756 int count = 0; 757 758 if (!p || p == current || p->state == TASK_RUNNING) 759 return 0; 760 761 /* 762 * Note: p may not be a blocked task (it could be current or 763 * another process running on some other CPU. Rather than 764 * trying to determine if p is really blocked, we just assume 765 * it's blocked and rely on the unwind routines to fail 766 * gracefully if the process wasn't really blocked after all. 767 * --davidm 99/12/15 768 */ 769 unw_init_from_blocked_task(&info, p); 770 do { 771 if (p->state == TASK_RUNNING) 772 return 0; 773 if (unw_unwind(&info) < 0) 774 return 0; 775 unw_get_ip(&info, &ip); 776 if (!in_sched_functions(ip)) 777 return ip; 778 } while (count++ < 16); 779 return 0; 780 } 781 782 void 783 cpu_halt (void) 784 { 785 pal_power_mgmt_info_u_t power_info[8]; 786 unsigned long min_power; 787 int i, min_power_state; 788 789 if (ia64_pal_halt_info(power_info) != 0) 790 return; 791 792 min_power_state = 0; 793 min_power = power_info[0].pal_power_mgmt_info_s.power_consumption; 794 for (i = 1; i < 8; ++i) 795 if (power_info[i].pal_power_mgmt_info_s.im 796 && power_info[i].pal_power_mgmt_info_s.power_consumption < min_power) { 797 min_power = power_info[i].pal_power_mgmt_info_s.power_consumption; 798 min_power_state = i; 799 } 800 801 while (1) 802 ia64_pal_halt(min_power_state); 803 } 804 805 void machine_shutdown(void) 806 { 807 #ifdef CONFIG_HOTPLUG_CPU 808 int cpu; 809 810 for_each_online_cpu(cpu) { 811 if (cpu != smp_processor_id()) 812 cpu_down(cpu); 813 } 814 #endif 815 #ifdef CONFIG_KEXEC 816 kexec_disable_iosapic(); 817 #endif 818 } 819 820 void 821 machine_restart (char *restart_cmd) 822 { 823 (void) notify_die(DIE_MACHINE_RESTART, restart_cmd, NULL, 0, 0, 0); 824 (*efi.reset_system)(EFI_RESET_WARM, 0, 0, NULL); 825 } 826 827 void 828 machine_halt (void) 829 { 830 (void) notify_die(DIE_MACHINE_HALT, "", NULL, 0, 0, 0); 831 cpu_halt(); 832 } 833 834 void 835 machine_power_off (void) 836 { 837 if (pm_power_off) 838 pm_power_off(); 839 machine_halt(); 840 } 841 842