arch/ia64/kernel/process.c

   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
  32 #include <asm/cpu.h>
  33 #include <asm/delay.h>
  34 #include <asm/elf.h>
  35 #include <asm/ia32.h>
  36 #include <asm/irq.h>
  37 #include <asm/kexec.h>
  38 #include <asm/pgalloc.h>
  39 #include <asm/processor.h>
  40 #include <asm/sal.h>
  41 #include <asm/tlbflush.h>
  42 #include <asm/uaccess.h>
  43 #include <asm/unwind.h>
  44 #include <asm/user.h>
  45
  46 #include "entry.h"
  47
  48 #ifdef CONFIG_PERFMON
  49 # include <asm/perfmon.h>
  50 #endif
  51
  52 #include "sigframe.h"
  53
  54 void (*ia64_mark_idle)(int);
  55
  56 unsigned long boot_option_idle_override = 0;
  57 EXPORT_SYMBOL(boot_option_idle_override);
  58
  59 void
  60 ia64_do_show_stack (struct unw_frame_info *info, void *arg)
  61 {
  62         unsigned long ip, sp, bsp;
  63         char buf[128];                  /* don't make it so big that it overflows the stack! */
  64
  65         printk("\nCall Trace:\n");
  66         do {
  67                 unw_get_ip(info, &ip);
  68                 if (ip == 0)
  69                         break;
  70
  71                 unw_get_sp(info, &sp);
  72                 unw_get_bsp(info, &bsp);
  73                 snprintf(buf, sizeof(buf),
  74                          " [<%016lx>] %%s\n"
  75                          "                                sp=%016lx bsp=%016lx\n",
  76                          ip, sp, bsp);
  77                 print_symbol(buf, ip);
  78         } while (unw_unwind(info) >= 0);
  79 }
  80
  81 void
  82 show_stack (struct task_struct *task, unsigned long *sp)
  83 {
  84         if (!task)
  85                 unw_init_running(ia64_do_show_stack, NULL);
  86         else {
  87                 struct unw_frame_info info;
  88
  89                 unw_init_from_blocked_task(&info, task);
  90                 ia64_do_show_stack(&info, NULL);
  91         }
  92 }
  93
  94 void
  95 dump_stack (void)
  96 {
  97         show_stack(NULL, NULL);
  98 }
  99
 100 EXPORT_SYMBOL(dump_stack);
 101
 102 void
 103 show_regs (struct pt_regs *regs)
 104 {
 105         unsigned long ip = regs->cr_iip + ia64_psr(regs)->ri;
 106
 107         print_modules();
 108         printk("\nPid: %d, CPU %d, comm: %20s\n", task_pid_nr(current),
 109                         smp_processor_id(), current->comm);
 110         printk("psr : %016lx ifs : %016lx ip  : [<%016lx>]    %s (%s)\n",
 111                regs->cr_ipsr, regs->cr_ifs, ip, print_tainted(),
 112                init_utsname()->release);
 113         print_symbol("ip is at %s\n", ip);
 114         printk("unat: %016lx pfs : %016lx rsc : %016lx\n",
 115                regs->ar_unat, regs->ar_pfs, regs->ar_rsc);
 116         printk("rnat: %016lx bsps: %016lx pr  : %016lx\n",
 117                regs->ar_rnat, regs->ar_bspstore, regs->pr);
 118         printk("ldrs: %016lx ccv : %016lx fpsr: %016lx\n",
 119                regs->loadrs, regs->ar_ccv, regs->ar_fpsr);
 120         printk("csd : %016lx ssd : %016lx\n", regs->ar_csd, regs->ar_ssd);
 121         printk("b0  : %016lx b6  : %016lx b7  : %016lx\n", regs->b0, regs->b6, regs->b7);
 122         printk("f6  : %05lx%016lx f7  : %05lx%016lx\n",
 123                regs->f6.u.bits[1], regs->f6.u.bits[0],
 124                regs->f7.u.bits[1], regs->f7.u.bits[0]);
 125         printk("f8  : %05lx%016lx f9  : %05lx%016lx\n",
 126                regs->f8.u.bits[1], regs->f8.u.bits[0],
 127                regs->f9.u.bits[1], regs->f9.u.bits[0]);
 128         printk("f10 : %05lx%016lx f11 : %05lx%016lx\n",
 129                regs->f10.u.bits[1], regs->f10.u.bits[0],
 130                regs->f11.u.bits[1], regs->f11.u.bits[0]);
 131
 132         printk("r1  : %016lx r2  : %016lx r3  : %016lx\n", regs->r1, regs->r2, regs->r3);
 133         printk("r8  : %016lx r9  : %016lx r10 : %016lx\n", regs->r8, regs->r9, regs->r10);
 134         printk("r11 : %016lx r12 : %016lx r13 : %016lx\n", regs->r11, regs->r12, regs->r13);
 135         printk("r14 : %016lx r15 : %016lx r16 : %016lx\n", regs->r14, regs->r15, regs->r16);
 136         printk("r17 : %016lx r18 : %016lx r19 : %016lx\n", regs->r17, regs->r18, regs->r19);
 137         printk("r20 : %016lx r21 : %016lx r22 : %016lx\n", regs->r20, regs->r21, regs->r22);
 138         printk("r23 : %016lx r24 : %016lx r25 : %016lx\n", regs->r23, regs->r24, regs->r25);
 139         printk("r26 : %016lx r27 : %016lx r28 : %016lx\n", regs->r26, regs->r27, regs->r28);
 140         printk("r29 : %016lx r30 : %016lx r31 : %016lx\n", regs->r29, regs->r30, regs->r31);
 141
 142         if (user_mode(regs)) {
 143                 /* print the stacked registers */
 144                 unsigned long val, *bsp, ndirty;
 145                 int i, sof, is_nat = 0;
 146
 147                 sof = regs->cr_ifs & 0x7f;      /* size of frame */
 148                 ndirty = (regs->loadrs >> 19);
 149                 bsp = ia64_rse_skip_regs((unsigned long *) regs->ar_bspstore, ndirty);
 150                 for (i = 0; i < sof; ++i) {
 151                         get_user(val, (unsigned long __user *) ia64_rse_skip_regs(bsp, i));
 152                         printk("r%-3u:%c%016lx%s", 32 + i, is_nat ? '*' : ' ', val,
 153                                ((i == sof - 1) || (i % 3) == 2) ? "\n" : " ");
 154                 }
 155         } else
 156                 show_stack(NULL, NULL);
 157 }
 158
 159 void tsk_clear_notify_resume(struct task_struct *tsk)
 160 {
 161 #ifdef CONFIG_PERFMON
 162         if (tsk->thread.pfm_needs_checking)
 163                 return;
 164 #endif
 165         if (test_ti_thread_flag(task_thread_info(tsk), TIF_RESTORE_RSE))
 166                 return;
 167         clear_ti_thread_flag(task_thread_info(tsk), TIF_NOTIFY_RESUME);
 168 }
 169
 170 /*
 171  * do_notify_resume_user():
 172  *      Called from notify_resume_user at entry.S, with interrupts disabled.
 173  */
 174 void
 175 do_notify_resume_user(sigset_t *unused, struct sigscratch *scr, long in_syscall)
 176 {
 177         if (fsys_mode(current, &scr->pt)) {
 178                 /*
 179                  * defer signal-handling etc. until we return to
 180                  * privilege-level 0.
 181                  */
 182                 if (!ia64_psr(&scr->pt)->lp)
 183                         ia64_psr(&scr->pt)->lp = 1;
 184                 return;
 185         }
 186
 187 #ifdef CONFIG_PERFMON
 188         if (current->thread.pfm_needs_checking)
 189                 /*
 190                  * Note: pfm_handle_work() allow us to call it with interrupts
 191                  * disabled, and may enable interrupts within the function.
 192                  */
 193                 pfm_handle_work();
 194 #endif
 195
 196         /* deal with pending signal delivery */
 197         if (test_thread_flag(TIF_SIGPENDING)) {
 198                 local_irq_enable();     /* force interrupt enable */
 199                 ia64_do_signal(scr, in_syscall);
 200         }
 201
 202         /* copy user rbs to kernel rbs */
 203         if (unlikely(test_thread_flag(TIF_RESTORE_RSE))) {
 204                 local_irq_enable();     /* force interrupt enable */
 205                 ia64_sync_krbs();
 206         }
 207
 208         local_irq_disable();    /* force interrupt disable */
 209 }
 210
 211 static int pal_halt        = 1;
 212 static int can_do_pal_halt = 1;
 213
 214 static int __init nohalt_setup(char * str)
 215 {
 216         pal_halt = can_do_pal_halt = 0;
 217         return 1;
 218 }
 219 __setup("nohalt", nohalt_setup);
 220
 221 void
 222 update_pal_halt_status(int status)
 223 {
 224         can_do_pal_halt = pal_halt && status;
 225 }
 226
 227 /*
 228  * We use this if we don't have any better idle routine..
 229  */
 230 void
 231 default_idle (void)
 232 {
 233         local_irq_enable();
 234         while (!need_resched()) {
 235                 if (can_do_pal_halt) {
 236                         local_irq_disable();
 237                         if (!need_resched()) {
 238                                 safe_halt();
 239                         }
 240                         local_irq_enable();
 241                 } else
 242                         cpu_relax();
 243         }
 244 }
 245
 246 #ifdef CONFIG_HOTPLUG_CPU
 247 /* We don't actually take CPU down, just spin without interrupts. */
 248 static inline void play_dead(void)
 249 {
 250         extern void ia64_cpu_local_tick (void);
 251         unsigned int this_cpu = smp_processor_id();
 252
 253         /* Ack it */
 254         __get_cpu_var(cpu_state) = CPU_DEAD;
 255
 256         max_xtp();
 257         local_irq_disable();
 258         idle_task_exit();
 259         ia64_jump_to_sal(&sal_boot_rendez_state[this_cpu]);
 260         /*
 261          * The above is a point of no-return, the processor is
 262          * expected to be in SAL loop now.
 263          */
 264         BUG();
 265 }
 266 #else
 267 static inline void play_dead(void)
 268 {
 269         BUG();
 270 }
 271 #endif /* CONFIG_HOTPLUG_CPU */
 272
 273 static void do_nothing(void *unused)
 274 {
 275 }
 276
 277 /*
 278  * cpu_idle_wait - Used to ensure that all the CPUs discard old value of
 279  * pm_idle and update to new pm_idle value. Required while changing pm_idle
 280  * handler on SMP systems.
 281  *
 282  * Caller must have changed pm_idle to the new value before the call. Old
 283  * pm_idle value will not be used by any CPU after the return of this function.
 284  */
 285 void cpu_idle_wait(void)
 286 {
 287         smp_mb();
 288         /* kick all the CPUs so that they exit out of pm_idle */
 289         smp_call_function(do_nothing, NULL, 0, 1);
 290 }
 291 EXPORT_SYMBOL_GPL(cpu_idle_wait);
 292
 293 void __attribute__((noreturn))
 294 cpu_idle (void)
 295 {
 296         void (*mark_idle)(int) = ia64_mark_idle;
 297         int cpu = smp_processor_id();
 298
 299         /* endless idle loop with no priority at all */
 300         while (1) {
 301                 if (can_do_pal_halt) {
 302                         current_thread_info()->status &= ~TS_POLLING;
 303                         /*
 304                          * TS_POLLING-cleared state must be visible before we
 305                          * test NEED_RESCHED:
 306                          */
 307                         smp_mb();
 308                 } else {
 309                         current_thread_info()->status |= TS_POLLING;
 310                 }
 311
 312                 if (!need_resched()) {
 313                         void (*idle)(void);
 314 #ifdef CONFIG_SMP
 315                         min_xtp();
 316 #endif
 317                         rmb();
 318                         if (mark_idle)
 319                                 (*mark_idle)(1);
 320
 321                         idle = pm_idle;
 322                         if (!idle)
 323                                 idle = default_idle;
 324                         (*idle)();
 325                         if (mark_idle)
 326                                 (*mark_idle)(0);
 327 #ifdef CONFIG_SMP
 328                         normal_xtp();
 329 #endif
 330                 }
 331                 preempt_enable_no_resched();
 332                 schedule();
 333                 preempt_disable();
 334                 check_pgt_cache();
 335                 if (cpu_is_offline(cpu))
 336                         play_dead();
 337         }
 338 }
 339
 340 void
 341 ia64_save_extra (struct task_struct *task)
 342 {
 343 #ifdef CONFIG_PERFMON
 344         unsigned long info;
 345 #endif
 346
 347         if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
 348                 ia64_save_debug_regs(&task->thread.dbr[0]);
 349
 350 #ifdef CONFIG_PERFMON
 351         if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
 352                 pfm_save_regs(task);
 353
 354         info = __get_cpu_var(pfm_syst_info);
 355         if (info & PFM_CPUINFO_SYST_WIDE)
 356                 pfm_syst_wide_update_task(task, info, 0);
 357 #endif
 358
 359 #ifdef CONFIG_IA32_SUPPORT
 360         if (IS_IA32_PROCESS(task_pt_regs(task)))
 361                 ia32_save_state(task);
 362 #endif
 363 }
 364
 365 void
 366 ia64_load_extra (struct task_struct *task)
 367 {
 368 #ifdef CONFIG_PERFMON
 369         unsigned long info;
 370 #endif
 371
 372         if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
 373                 ia64_load_debug_regs(&task->thread.dbr[0]);
 374
 375 #ifdef CONFIG_PERFMON
 376         if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
 377                 pfm_load_regs(task);
 378
 379         info = __get_cpu_var(pfm_syst_info);
 380         if (info & PFM_CPUINFO_SYST_WIDE)
 381                 pfm_syst_wide_update_task(task, info, 1);
 382 #endif
 383
 384 #ifdef CONFIG_IA32_SUPPORT
 385         if (IS_IA32_PROCESS(task_pt_regs(task)))
 386                 ia32_load_state(task);
 387 #endif
 388 }
 389
 390 /*
 391  * Copy the state of an ia-64 thread.
 392  *
 393  * We get here through the following  call chain:
 394  *
 395  *      from user-level:        from kernel:
 396  *
 397  *      <clone syscall>         <some kernel call frames>
 398  *      sys_clone                  :
 399  *      do_fork                 do_fork
 400  *      copy_thread             copy_thread
 401  *
 402  * This means that the stack layout is as follows:
 403  *
 404  *      +---------------------+ (highest addr)
 405  *      |   struct pt_regs    |
 406  *      +---------------------+
 407  *      | struct switch_stack |
 408  *      +---------------------+
 409  *      |                     |
 410  *      |    memory stack     |
 411  *      |                     | <-- sp (lowest addr)
 412  *      +---------------------+
 413  *
 414  * Observe that we copy the unat values that are in pt_regs and switch_stack.  Spilling an
 415  * integer to address X causes bit N in ar.unat to be set to the NaT bit of the register,
 416  * with N=(X & 0x1ff)/8.  Thus, copying the unat value preserves the NaT bits ONLY if the
 417  * pt_regs structure in the parent is congruent to that of the child, modulo 512.  Since
 418  * the stack is page aligned and the page size is at least 4KB, this is always the case,
 419  * so there is nothing to worry about.
 420  */
 421 int
 422 copy_thread (int nr, unsigned long clone_flags,
 423              unsigned long user_stack_base, unsigned long user_stack_size,
 424              struct task_struct *p, struct pt_regs *regs)
 425 {
 426         extern char ia64_ret_from_clone, ia32_ret_from_clone;
 427         struct switch_stack *child_stack, *stack;
 428         unsigned long rbs, child_rbs, rbs_size;
 429         struct pt_regs *child_ptregs;
 430         int retval = 0;
 431
 432 #ifdef CONFIG_SMP
 433         /*
 434          * For SMP idle threads, fork_by_hand() calls do_fork with
 435          * NULL regs.
 436          */
 437         if (!regs)
 438                 return 0;
 439 #endif
 440
 441         stack = ((struct switch_stack *) regs) - 1;
 442
 443         child_ptregs = (struct pt_regs *) ((unsigned long) p + IA64_STK_OFFSET) - 1;
 444         child_stack = (struct switch_stack *) child_ptregs - 1;
 445
 446         /* copy parent's switch_stack & pt_regs to child: */
 447         memcpy(child_stack, stack, sizeof(*child_ptregs) + sizeof(*child_stack));
 448
 449         rbs = (unsigned long) current + IA64_RBS_OFFSET;
 450         child_rbs = (unsigned long) p + IA64_RBS_OFFSET;
 451         rbs_size = stack->ar_bspstore - rbs;
 452
 453         /* copy the parent's register backing store to the child: */
 454         memcpy((void *) child_rbs, (void *) rbs, rbs_size);
 455
 456         if (likely(user_mode(child_ptregs))) {
 457                 if ((clone_flags & CLONE_SETTLS) && !IS_IA32_PROCESS(regs))
 458                         child_ptregs->r13 = regs->r16;  /* see sys_clone2() in entry.S */
 459                 if (user_stack_base) {
 460                         child_ptregs->r12 = user_stack_base + user_stack_size - 16;
 461                         child_ptregs->ar_bspstore = user_stack_base;
 462                         child_ptregs->ar_rnat = 0;
 463                         child_ptregs->loadrs = 0;
 464                 }
 465         } else {
 466                 /*
 467                  * Note: we simply preserve the relative position of
 468                  * the stack pointer here.  There is no need to
 469                  * allocate a scratch area here, since that will have
 470                  * been taken care of by the caller of sys_clone()
 471                  * already.
 472                  */
 473                 child_ptregs->r12 = (unsigned long) child_ptregs - 16; /* kernel sp */
 474                 child_ptregs->r13 = (unsigned long) p;          /* set `current' pointer */
 475         }
 476         child_stack->ar_bspstore = child_rbs + rbs_size;
 477         if (IS_IA32_PROCESS(regs))
 478                 child_stack->b0 = (unsigned long) &ia32_ret_from_clone;
 479         else
 480                 child_stack->b0 = (unsigned long) &ia64_ret_from_clone;
 481
 482         /* copy parts of thread_struct: */
 483         p->thread.ksp = (unsigned long) child_stack - 16;
 484
 485         /* stop some PSR bits from being inherited.
 486          * the psr.up/psr.pp bits must be cleared on fork but inherited on execve()
 487          * therefore we must specify them explicitly here and not include them in
 488          * IA64_PSR_BITS_TO_CLEAR.
 489          */
 490         child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET)
 491                                  & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP));
 492
 493         /*
 494          * NOTE: The calling convention considers all floating point
 495          * registers in the high partition (fph) to be scratch.  Since
 496          * the only way to get to this point is through a system call,
 497          * we know that the values in fph are all dead.  Hence, there
 498          * is no need to inherit the fph state from the parent to the
 499          * child and all we have to do is to make sure that
 500          * IA64_THREAD_FPH_VALID is cleared in the child.
 501          *
 502          * XXX We could push this optimization a bit further by
 503          * clearing IA64_THREAD_FPH_VALID on ANY system call.
 504          * However, it's not clear this is worth doing.  Also, it
 505          * would be a slight deviation from the normal Linux system
 506          * call behavior where scratch registers are preserved across
 507          * system calls (unless used by the system call itself).
 508          */
 509 #       define THREAD_FLAGS_TO_CLEAR    (IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID \
 510                                          | IA64_THREAD_PM_VALID)
 511 #       define THREAD_FLAGS_TO_SET      0
 512         p->thread.flags = ((current->thread.flags & ~THREAD_FLAGS_TO_CLEAR)
 513                            | THREAD_FLAGS_TO_SET);
 514         ia64_drop_fpu(p);       /* don't pick up stale state from a CPU's fph */
 515 #ifdef CONFIG_IA32_SUPPORT
 516         /*
 517          * If we're cloning an IA32 task then save the IA32 extra
 518          * state from the current task to the new task
 519          */
 520         if (IS_IA32_PROCESS(task_pt_regs(current))) {
 521                 ia32_save_state(p);
 522                 if (clone_flags & CLONE_SETTLS)
 523                         retval = ia32_clone_tls(p, child_ptregs);
 524
 525                 /* Copy partially mapped page list */
 526                 if (!retval)
 527                         retval = ia32_copy_ia64_partial_page_list(p,
 528                                                                 clone_flags);
 529         }
 530 #endif
 531
 532 #ifdef CONFIG_PERFMON
 533         if (current->thread.pfm_context)
 534                 pfm_inherit(p, child_ptregs);
 535 #endif
 536         return retval;
 537 }
 538
 539 static void
 540 do_copy_task_regs (struct task_struct *task, struct unw_frame_info *info, void *arg)
 541 {
 542         unsigned long mask, sp, nat_bits = 0, ar_rnat, urbs_end, cfm;
 543         unsigned long uninitialized_var(ip);    /* GCC be quiet */
 544         elf_greg_t *dst = arg;
 545         struct pt_regs *pt;
 546         char nat;
 547         int i;
 548
 549         memset(dst, 0, sizeof(elf_gregset_t));  /* don't leak any kernel bits to user-level */
 550
 551         if (unw_unwind_to_user(info) < 0)
 552                 return;
 553
 554         unw_get_sp(info, &sp);
 555         pt = (struct pt_regs *) (sp + 16);
 556
 557         urbs_end = ia64_get_user_rbs_end(task, pt, &cfm);
 558
 559         if (ia64_sync_user_rbs(task, info->sw, pt->ar_bspstore, urbs_end) < 0)
 560                 return;
 561
 562         ia64_peek(task, info->sw, urbs_end, (long) ia64_rse_rnat_addr((long *) urbs_end),
 563                   &ar_rnat);
 564
 565         /*
 566          * coredump format:
 567          *      r0-r31
 568          *      NaT bits (for r0-r31; bit N == 1 iff rN is a NaT)
 569          *      predicate registers (p0-p63)
 570          *      b0-b7
 571          *      ip cfm user-mask
 572          *      ar.rsc ar.bsp ar.bspstore ar.rnat
 573          *      ar.ccv ar.unat ar.fpsr ar.pfs ar.lc ar.ec
 574          */
 575
 576         /* r0 is zero */
 577         for (i = 1, mask = (1UL << i); i < 32; ++i) {
 578                 unw_get_gr(info, i, &dst[i], &nat);
 579                 if (nat)
 580                         nat_bits |= mask;
 581                 mask <<= 1;
 582         }
 583         dst[32] = nat_bits;
 584         unw_get_pr(info, &dst[33]);
 585
 586         for (i = 0; i < 8; ++i)
 587                 unw_get_br(info, i, &dst[34 + i]);
 588
 589         unw_get_rp(info, &ip);
 590         dst[42] = ip + ia64_psr(pt)->ri;
 591         dst[43] = cfm;
 592         dst[44] = pt->cr_ipsr & IA64_PSR_UM;
 593
 594         unw_get_ar(info, UNW_AR_RSC, &dst[45]);
 595         /*
 596          * For bsp and bspstore, unw_get_ar() would return the kernel
 597          * addresses, but we need the user-level addresses instead:
 598          */
 599         dst[46] = urbs_end;     /* note: by convention PT_AR_BSP points to the end of the urbs! */
 600         dst[47] = pt->ar_bspstore;
 601         dst[48] = ar_rnat;
 602         unw_get_ar(info, UNW_AR_CCV, &dst[49]);
 603         unw_get_ar(info, UNW_AR_UNAT, &dst[50]);
 604         unw_get_ar(info, UNW_AR_FPSR, &dst[51]);
 605         dst[52] = pt->ar_pfs;   /* UNW_AR_PFS is == to pt->cr_ifs for interrupt frames */
 606         unw_get_ar(info, UNW_AR_LC, &dst[53]);
 607         unw_get_ar(info, UNW_AR_EC, &dst[54]);
 608         unw_get_ar(info, UNW_AR_CSD, &dst[55]);
 609         unw_get_ar(info, UNW_AR_SSD, &dst[56]);
 610 }
 611
 612 void
 613 do_dump_task_fpu (struct task_struct *task, struct unw_frame_info *info, void *arg)
 614 {
 615         elf_fpreg_t *dst = arg;
 616         int i;
 617
 618         memset(dst, 0, sizeof(elf_fpregset_t)); /* don't leak any "random" bits */
 619
 620         if (unw_unwind_to_user(info) < 0)
 621                 return;
 622
 623         /* f0 is 0.0, f1 is 1.0 */
 624
 625         for (i = 2; i < 32; ++i)
 626                 unw_get_fr(info, i, dst + i);
 627
 628         ia64_flush_fph(task);
 629         if ((task->thread.flags & IA64_THREAD_FPH_VALID) != 0)
 630                 memcpy(dst + 32, task->thread.fph, 96*16);
 631 }
 632
 633 void
 634 do_copy_regs (struct unw_frame_info *info, void *arg)
 635 {
 636         do_copy_task_regs(current, info, arg);
 637 }
 638
 639 void
 640 do_dump_fpu (struct unw_frame_info *info, void *arg)
 641 {
 642         do_dump_task_fpu(current, info, arg);
 643 }
 644
 645 void
 646 ia64_elf_core_copy_regs (struct pt_regs *pt, elf_gregset_t dst)
 647 {
 648         unw_init_running(do_copy_regs, dst);
 649 }
 650
 651 int
 652 dump_fpu (struct pt_regs *pt, elf_fpregset_t dst)
 653 {
 654         unw_init_running(do_dump_fpu, dst);
 655         return 1;       /* f0-f31 are always valid so we always return 1 */
 656 }
 657
 658 long
 659 sys_execve (char __user *filename, char __user * __user *argv, char __user * __user *envp,
 660             struct pt_regs *regs)
 661 {
 662         char *fname;
 663         int error;
 664
 665         fname = getname(filename);
 666         error = PTR_ERR(fname);
 667         if (IS_ERR(fname))
 668                 goto out;
 669         error = do_execve(fname, argv, envp, regs);
 670         putname(fname);
 671 out:
 672         return error;
 673 }
 674
 675 pid_t
 676 kernel_thread (int (*fn)(void *), void *arg, unsigned long flags)
 677 {
 678         extern void start_kernel_thread (void);
 679         unsigned long *helper_fptr = (unsigned long *) &start_kernel_thread;
 680         struct {
 681                 struct switch_stack sw;
 682                 struct pt_regs pt;
 683         } regs;
 684
 685         memset(&regs, 0, sizeof(regs));
 686         regs.pt.cr_iip = helper_fptr[0];        /* set entry point (IP) */
 687         regs.pt.r1 = helper_fptr[1];            /* set GP */
 688         regs.pt.r9 = (unsigned long) fn;        /* 1st argument */
 689         regs.pt.r11 = (unsigned long) arg;      /* 2nd argument */
 690         /* Preserve PSR bits, except for bits 32-34 and 37-45, which we can't read.  */
 691         regs.pt.cr_ipsr = ia64_getreg(_IA64_REG_PSR) | IA64_PSR_BN;
 692         regs.pt.cr_ifs = 1UL << 63;             /* mark as valid, empty frame */
 693         regs.sw.ar_fpsr = regs.pt.ar_fpsr = ia64_getreg(_IA64_REG_AR_FPSR);
 694         regs.sw.ar_bspstore = (unsigned long) current + IA64_RBS_OFFSET;
 695         regs.sw.pr = (1 << PRED_KERNEL_STACK);
 696         return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, &regs.pt, 0, NULL, NULL);
 697 }
 698 EXPORT_SYMBOL(kernel_thread);
 699
 700 /* This gets called from kernel_thread() via ia64_invoke_thread_helper().  */
 701 int
 702 kernel_thread_helper (int (*fn)(void *), void *arg)
 703 {
 704 #ifdef CONFIG_IA32_SUPPORT
 705         if (IS_IA32_PROCESS(task_pt_regs(current))) {
 706                 /* A kernel thread is always a 64-bit process. */
 707                 current->thread.map_base  = DEFAULT_MAP_BASE;
 708                 current->thread.task_size = DEFAULT_TASK_SIZE;
 709                 ia64_set_kr(IA64_KR_IO_BASE, current->thread.old_iob);
 710                 ia64_set_kr(IA64_KR_TSSD, current->thread.old_k1);
 711         }
 712 #endif
 713         return (*fn)(arg);
 714 }
 715
 716 /*
 717  * Flush thread state.  This is called when a thread does an execve().
 718  */
 719 void
 720 flush_thread (void)
 721 {
 722         /* drop floating-point and debug-register state if it exists: */
 723         current->thread.flags &= ~(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID);
 724         ia64_drop_fpu(current);
 725 #ifdef CONFIG_IA32_SUPPORT
 726         if (IS_IA32_PROCESS(task_pt_regs(current))) {
 727                 ia32_drop_ia64_partial_page_list(current);
 728                 current->thread.task_size = IA32_PAGE_OFFSET;
 729                 set_fs(USER_DS);
 730                 memset(current->thread.tls_array, 0, sizeof(current->thread.tls_array));
 731         }
 732 #endif
 733 }
 734
 735 /*
 736  * Clean up state associated with current thread.  This is called when
 737  * the thread calls exit().
 738  */
 739 void
 740 exit_thread (void)
 741 {
 742
 743         ia64_drop_fpu(current);
 744 #ifdef CONFIG_PERFMON
 745        /* if needed, stop monitoring and flush state to perfmon context */
 746         if (current->thread.pfm_context)
 747                 pfm_exit_thread(current);
 748
 749         /* free debug register resources */
 750         if (current->thread.flags & IA64_THREAD_DBG_VALID)
 751                 pfm_release_debug_registers(current);
 752 #endif
 753         if (IS_IA32_PROCESS(task_pt_regs(current)))
 754                 ia32_drop_ia64_partial_page_list(current);
 755 }
 756
 757 unsigned long
 758 get_wchan (struct task_struct *p)
 759 {
 760         struct unw_frame_info info;
 761         unsigned long ip;
 762         int count = 0;
 763
 764         if (!p || p == current || p->state == TASK_RUNNING)
 765                 return 0;
 766
 767         /*
 768          * Note: p may not be a blocked task (it could be current or
 769          * another process running on some other CPU.  Rather than
 770          * trying to determine if p is really blocked, we just assume
 771          * it's blocked and rely on the unwind routines to fail
 772          * gracefully if the process wasn't really blocked after all.
 773          * --davidm 99/12/15
 774          */
 775         unw_init_from_blocked_task(&info, p);
 776         do {
 777                 if (p->state == TASK_RUNNING)
 778                         return 0;
 779                 if (unw_unwind(&info) < 0)
 780                         return 0;
 781                 unw_get_ip(&info, &ip);
 782                 if (!in_sched_functions(ip))
 783                         return ip;
 784         } while (count++ < 16);
 785         return 0;
 786 }
 787
 788 void
 789 cpu_halt (void)
 790 {
 791         pal_power_mgmt_info_u_t power_info[8];
 792         unsigned long min_power;
 793         int i, min_power_state;
 794
 795         if (ia64_pal_halt_info(power_info) != 0)
 796                 return;
 797
 798         min_power_state = 0;
 799         min_power = power_info[0].pal_power_mgmt_info_s.power_consumption;
 800         for (i = 1; i < 8; ++i)
 801                 if (power_info[i].pal_power_mgmt_info_s.im
 802                     && power_info[i].pal_power_mgmt_info_s.power_consumption < min_power) {
 803                         min_power = power_info[i].pal_power_mgmt_info_s.power_consumption;
 804                         min_power_state = i;
 805                 }
 806
 807         while (1)
 808                 ia64_pal_halt(min_power_state);
 809 }
 810
 811 void machine_shutdown(void)
 812 {
 813 #ifdef CONFIG_HOTPLUG_CPU
 814         int cpu;
 815
 816         for_each_online_cpu(cpu) {
 817                 if (cpu != smp_processor_id())
 818                         cpu_down(cpu);
 819         }
 820 #endif
 821 #ifdef CONFIG_KEXEC
 822         kexec_disable_iosapic();
 823 #endif
 824 }
 825
 826 void
 827 machine_restart (char *restart_cmd)
 828 {
 829         (void) notify_die(DIE_MACHINE_RESTART, restart_cmd, NULL, 0, 0, 0);
 830         (*efi.reset_system)(EFI_RESET_WARM, 0, 0, NULL);
 831 }
 832
 833 void
 834 machine_halt (void)
 835 {
 836         (void) notify_die(DIE_MACHINE_HALT, "", NULL, 0, 0, 0);
 837         cpu_halt();
 838 }
 839
 840 void
 841 machine_power_off (void)
 842 {
 843         if (pm_power_off)
 844                 pm_power_off();
 845         machine_halt();
 846 }
 847