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 #define __KERNEL_SYSCALLS__     /* see <asm/unistd.h> */
  12 #include <linux/config.h>
  13
  14 #include <linux/cpu.h>
  15 #include <linux/pm.h>
  16 #include <linux/elf.h>
  17 #include <linux/errno.h>
  18 #include <linux/kallsyms.h>
  19 #include <linux/kernel.h>
  20 #include <linux/mm.h>
  21 #include <linux/module.h>
  22 #include <linux/notifier.h>
  23 #include <linux/personality.h>
  24 #include <linux/sched.h>
  25 #include <linux/slab.h>
  26 #include <linux/smp_lock.h>
  27 #include <linux/stddef.h>
  28 #include <linux/thread_info.h>
  29 #include <linux/unistd.h>
  30 #include <linux/efi.h>
  31 #include <linux/interrupt.h>
  32 #include <linux/delay.h>
  33 #include <linux/kprobes.h>
  34
  35 #include <asm/cpu.h>
  36 #include <asm/delay.h>
  37 #include <asm/elf.h>
  38 #include <asm/ia32.h>
  39 #include <asm/irq.h>
  40 #include <asm/kdebug.h>
  41 #include <asm/pgalloc.h>
  42 #include <asm/processor.h>
  43 #include <asm/sal.h>
  44 #include <asm/tlbflush.h>
  45 #include <asm/uaccess.h>
  46 #include <asm/unwind.h>
  47 #include <asm/user.h>
  48
  49 #include "entry.h"
  50
  51 #ifdef CONFIG_PERFMON
  52 # include <asm/perfmon.h>
  53 #endif
  54
  55 #include "sigframe.h"
  56
  57 void (*ia64_mark_idle)(int);
  58 static DEFINE_PER_CPU(unsigned int, cpu_idle_state);
  59
  60 unsigned long boot_option_idle_override = 0;
  61 EXPORT_SYMBOL(boot_option_idle_override);
  62
  63 void
  64 ia64_do_show_stack (struct unw_frame_info *info, void *arg)
  65 {
  66         unsigned long ip, sp, bsp;
  67         char buf[128];                  /* don't make it so big that it overflows the stack! */
  68
  69         printk("\nCall Trace:\n");
  70         do {
  71                 unw_get_ip(info, &ip);
  72                 if (ip == 0)
  73                         break;
  74
  75                 unw_get_sp(info, &sp);
  76                 unw_get_bsp(info, &bsp);
  77                 snprintf(buf, sizeof(buf),
  78                          " [<%016lx>] %%s\n"
  79                          "                                sp=%016lx bsp=%016lx\n",
  80                          ip, sp, bsp);
  81                 print_symbol(buf, ip);
  82         } while (unw_unwind(info) >= 0);
  83 }
  84
  85 void
  86 show_stack (struct task_struct *task, unsigned long *sp)
  87 {
  88         if (!task)
  89                 unw_init_running(ia64_do_show_stack, NULL);
  90         else {
  91                 struct unw_frame_info info;
  92
  93                 unw_init_from_blocked_task(&info, task);
  94                 ia64_do_show_stack(&info, NULL);
  95         }
  96 }
  97
  98 void
  99 dump_stack (void)
 100 {
 101         show_stack(NULL, NULL);
 102 }
 103
 104 EXPORT_SYMBOL(dump_stack);
 105
 106 void
 107 show_regs (struct pt_regs *regs)
 108 {
 109         unsigned long ip = regs->cr_iip + ia64_psr(regs)->ri;
 110
 111         print_modules();
 112         printk("\nPid: %d, CPU %d, comm: %20s\n", current->pid, smp_processor_id(), current->comm);
 113         printk("psr : %016lx ifs : %016lx ip  : [<%016lx>]    %s\n",
 114                regs->cr_ipsr, regs->cr_ifs, ip, print_tainted());
 115         print_symbol("ip is at %s\n", ip);
 116         printk("unat: %016lx pfs : %016lx rsc : %016lx\n",
 117                regs->ar_unat, regs->ar_pfs, regs->ar_rsc);
 118         printk("rnat: %016lx bsps: %016lx pr  : %016lx\n",
 119                regs->ar_rnat, regs->ar_bspstore, regs->pr);
 120         printk("ldrs: %016lx ccv : %016lx fpsr: %016lx\n",
 121                regs->loadrs, regs->ar_ccv, regs->ar_fpsr);
 122         printk("csd : %016lx ssd : %016lx\n", regs->ar_csd, regs->ar_ssd);
 123         printk("b0  : %016lx b6  : %016lx b7  : %016lx\n", regs->b0, regs->b6, regs->b7);
 124         printk("f6  : %05lx%016lx f7  : %05lx%016lx\n",
 125                regs->f6.u.bits[1], regs->f6.u.bits[0],
 126                regs->f7.u.bits[1], regs->f7.u.bits[0]);
 127         printk("f8  : %05lx%016lx f9  : %05lx%016lx\n",
 128                regs->f8.u.bits[1], regs->f8.u.bits[0],
 129                regs->f9.u.bits[1], regs->f9.u.bits[0]);
 130         printk("f10 : %05lx%016lx f11 : %05lx%016lx\n",
 131                regs->f10.u.bits[1], regs->f10.u.bits[0],
 132                regs->f11.u.bits[1], regs->f11.u.bits[0]);
 133
 134         printk("r1  : %016lx r2  : %016lx r3  : %016lx\n", regs->r1, regs->r2, regs->r3);
 135         printk("r8  : %016lx r9  : %016lx r10 : %016lx\n", regs->r8, regs->r9, regs->r10);
 136         printk("r11 : %016lx r12 : %016lx r13 : %016lx\n", regs->r11, regs->r12, regs->r13);
 137         printk("r14 : %016lx r15 : %016lx r16 : %016lx\n", regs->r14, regs->r15, regs->r16);
 138         printk("r17 : %016lx r18 : %016lx r19 : %016lx\n", regs->r17, regs->r18, regs->r19);
 139         printk("r20 : %016lx r21 : %016lx r22 : %016lx\n", regs->r20, regs->r21, regs->r22);
 140         printk("r23 : %016lx r24 : %016lx r25 : %016lx\n", regs->r23, regs->r24, regs->r25);
 141         printk("r26 : %016lx r27 : %016lx r28 : %016lx\n", regs->r26, regs->r27, regs->r28);
 142         printk("r29 : %016lx r30 : %016lx r31 : %016lx\n", regs->r29, regs->r30, regs->r31);
 143
 144         if (user_mode(regs)) {
 145                 /* print the stacked registers */
 146                 unsigned long val, *bsp, ndirty;
 147                 int i, sof, is_nat = 0;
 148
 149                 sof = regs->cr_ifs & 0x7f;      /* size of frame */
 150                 ndirty = (regs->loadrs >> 19);
 151                 bsp = ia64_rse_skip_regs((unsigned long *) regs->ar_bspstore, ndirty);
 152                 for (i = 0; i < sof; ++i) {
 153                         get_user(val, (unsigned long __user *) ia64_rse_skip_regs(bsp, i));
 154                         printk("r%-3u:%c%016lx%s", 32 + i, is_nat ? '*' : ' ', val,
 155                                ((i == sof - 1) || (i % 3) == 2) ? "\n" : " ");
 156                 }
 157         } else
 158                 show_stack(NULL, NULL);
 159 }
 160
 161 void
 162 do_notify_resume_user (sigset_t *oldset, struct sigscratch *scr, long in_syscall)
 163 {
 164         if (fsys_mode(current, &scr->pt)) {
 165                 /* defer signal-handling etc. until we return to privilege-level 0.  */
 166                 if (!ia64_psr(&scr->pt)->lp)
 167                         ia64_psr(&scr->pt)->lp = 1;
 168                 return;
 169         }
 170
 171 #ifdef CONFIG_PERFMON
 172         if (current->thread.pfm_needs_checking)
 173                 pfm_handle_work();
 174 #endif
 175
 176         /* deal with pending signal delivery */
 177         if (test_thread_flag(TIF_SIGPENDING))
 178                 ia64_do_signal(oldset, scr, in_syscall);
 179 }
 180
 181 static int pal_halt        = 1;
 182 static int can_do_pal_halt = 1;
 183
 184 static int __init nohalt_setup(char * str)
 185 {
 186         pal_halt = can_do_pal_halt = 0;
 187         return 1;
 188 }
 189 __setup("nohalt", nohalt_setup);
 190
 191 void
 192 update_pal_halt_status(int status)
 193 {
 194         can_do_pal_halt = pal_halt && status;
 195 }
 196
 197 /*
 198  * We use this if we don't have any better idle routine..
 199  */
 200 void
 201 default_idle (void)
 202 {
 203         local_irq_enable();
 204         while (!need_resched()) {
 205                 if (can_do_pal_halt)
 206                         safe_halt();
 207                 else
 208                         cpu_relax();
 209         }
 210 }
 211
 212 #ifdef CONFIG_HOTPLUG_CPU
 213 /* We don't actually take CPU down, just spin without interrupts. */
 214 static inline void play_dead(void)
 215 {
 216         extern void ia64_cpu_local_tick (void);
 217         unsigned int this_cpu = smp_processor_id();
 218
 219         /* Ack it */
 220         __get_cpu_var(cpu_state) = CPU_DEAD;
 221
 222         max_xtp();
 223         local_irq_disable();
 224         idle_task_exit();
 225         ia64_jump_to_sal(&sal_boot_rendez_state[this_cpu]);
 226         /*
 227          * The above is a point of no-return, the processor is
 228          * expected to be in SAL loop now.
 229          */
 230         BUG();
 231 }
 232 #else
 233 static inline void play_dead(void)
 234 {
 235         BUG();
 236 }
 237 #endif /* CONFIG_HOTPLUG_CPU */
 238
 239 void cpu_idle_wait(void)
 240 {
 241         unsigned int cpu, this_cpu = get_cpu();
 242         cpumask_t map;
 243
 244         set_cpus_allowed(current, cpumask_of_cpu(this_cpu));
 245         put_cpu();
 246
 247         cpus_clear(map);
 248         for_each_online_cpu(cpu) {
 249                 per_cpu(cpu_idle_state, cpu) = 1;
 250                 cpu_set(cpu, map);
 251         }
 252
 253         __get_cpu_var(cpu_idle_state) = 0;
 254
 255         wmb();
 256         do {
 257                 ssleep(1);
 258                 for_each_online_cpu(cpu) {
 259                         if (cpu_isset(cpu, map) && !per_cpu(cpu_idle_state, cpu))
 260                                 cpu_clear(cpu, map);
 261                 }
 262                 cpus_and(map, map, cpu_online_map);
 263         } while (!cpus_empty(map));
 264 }
 265 EXPORT_SYMBOL_GPL(cpu_idle_wait);
 266
 267 void __attribute__((noreturn))
 268 cpu_idle (void)
 269 {
 270         void (*mark_idle)(int) = ia64_mark_idle;
 271         int cpu = smp_processor_id();
 272
 273         /* endless idle loop with no priority at all */
 274         while (1) {
 275                 if (can_do_pal_halt)
 276                         clear_thread_flag(TIF_POLLING_NRFLAG);
 277                 else
 278                         set_thread_flag(TIF_POLLING_NRFLAG);
 279
 280                 if (!need_resched()) {
 281                         void (*idle)(void);
 282 #ifdef CONFIG_SMP
 283                         min_xtp();
 284 #endif
 285                         if (__get_cpu_var(cpu_idle_state))
 286                                 __get_cpu_var(cpu_idle_state) = 0;
 287
 288                         rmb();
 289                         if (mark_idle)
 290                                 (*mark_idle)(1);
 291
 292                         idle = pm_idle;
 293                         if (!idle)
 294                                 idle = default_idle;
 295                         (*idle)();
 296                         if (mark_idle)
 297                                 (*mark_idle)(0);
 298 #ifdef CONFIG_SMP
 299                         normal_xtp();
 300 #endif
 301                 }
 302                 preempt_enable_no_resched();
 303                 schedule();
 304                 preempt_disable();
 305                 check_pgt_cache();
 306                 if (cpu_is_offline(cpu))
 307                         play_dead();
 308         }
 309 }
 310
 311 void
 312 ia64_save_extra (struct task_struct *task)
 313 {
 314 #ifdef CONFIG_PERFMON
 315         unsigned long info;
 316 #endif
 317
 318         if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
 319                 ia64_save_debug_regs(&task->thread.dbr[0]);
 320
 321 #ifdef CONFIG_PERFMON
 322         if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
 323                 pfm_save_regs(task);
 324
 325         info = __get_cpu_var(pfm_syst_info);
 326         if (info & PFM_CPUINFO_SYST_WIDE)
 327                 pfm_syst_wide_update_task(task, info, 0);
 328 #endif
 329
 330 #ifdef CONFIG_IA32_SUPPORT
 331         if (IS_IA32_PROCESS(task_pt_regs(task)))
 332                 ia32_save_state(task);
 333 #endif
 334 }
 335
 336 void
 337 ia64_load_extra (struct task_struct *task)
 338 {
 339 #ifdef CONFIG_PERFMON
 340         unsigned long info;
 341 #endif
 342
 343         if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
 344                 ia64_load_debug_regs(&task->thread.dbr[0]);
 345
 346 #ifdef CONFIG_PERFMON
 347         if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
 348                 pfm_load_regs(task);
 349
 350         info = __get_cpu_var(pfm_syst_info);
 351         if (info & PFM_CPUINFO_SYST_WIDE)
 352                 pfm_syst_wide_update_task(task, info, 1);
 353 #endif
 354
 355 #ifdef CONFIG_IA32_SUPPORT
 356         if (IS_IA32_PROCESS(task_pt_regs(task)))
 357                 ia32_load_state(task);
 358 #endif
 359 }
 360
 361 /*
 362  * Copy the state of an ia-64 thread.
 363  *
 364  * We get here through the following  call chain:
 365  *
 366  *      from user-level:        from kernel:
 367  *
 368  *      <clone syscall>         <some kernel call frames>
 369  *      sys_clone                  :
 370  *      do_fork                 do_fork
 371  *      copy_thread             copy_thread
 372  *
 373  * This means that the stack layout is as follows:
 374  *
 375  *      +---------------------+ (highest addr)
 376  *      |   struct pt_regs    |
 377  *      +---------------------+
 378  *      | struct switch_stack |
 379  *      +---------------------+
 380  *      |                     |
 381  *      |    memory stack     |
 382  *      |                     | <-- sp (lowest addr)
 383  *      +---------------------+
 384  *
 385  * Observe that we copy the unat values that are in pt_regs and switch_stack.  Spilling an
 386  * integer to address X causes bit N in ar.unat to be set to the NaT bit of the register,
 387  * with N=(X & 0x1ff)/8.  Thus, copying the unat value preserves the NaT bits ONLY if the
 388  * pt_regs structure in the parent is congruent to that of the child, modulo 512.  Since
 389  * the stack is page aligned and the page size is at least 4KB, this is always the case,
 390  * so there is nothing to worry about.
 391  */
 392 int
 393 copy_thread (int nr, unsigned long clone_flags,
 394              unsigned long user_stack_base, unsigned long user_stack_size,
 395              struct task_struct *p, struct pt_regs *regs)
 396 {
 397         extern char ia64_ret_from_clone, ia32_ret_from_clone;
 398         struct switch_stack *child_stack, *stack;
 399         unsigned long rbs, child_rbs, rbs_size;
 400         struct pt_regs *child_ptregs;
 401         int retval = 0;
 402
 403 #ifdef CONFIG_SMP
 404         /*
 405          * For SMP idle threads, fork_by_hand() calls do_fork with
 406          * NULL regs.
 407          */
 408         if (!regs)
 409                 return 0;
 410 #endif
 411
 412         stack = ((struct switch_stack *) regs) - 1;
 413
 414         child_ptregs = (struct pt_regs *) ((unsigned long) p + IA64_STK_OFFSET) - 1;
 415         child_stack = (struct switch_stack *) child_ptregs - 1;
 416
 417         /* copy parent's switch_stack & pt_regs to child: */
 418         memcpy(child_stack, stack, sizeof(*child_ptregs) + sizeof(*child_stack));
 419
 420         rbs = (unsigned long) current + IA64_RBS_OFFSET;
 421         child_rbs = (unsigned long) p + IA64_RBS_OFFSET;
 422         rbs_size = stack->ar_bspstore - rbs;
 423
 424         /* copy the parent's register backing store to the child: */
 425         memcpy((void *) child_rbs, (void *) rbs, rbs_size);
 426
 427         if (likely(user_mode(child_ptregs))) {
 428                 if ((clone_flags & CLONE_SETTLS) && !IS_IA32_PROCESS(regs))
 429                         child_ptregs->r13 = regs->r16;  /* see sys_clone2() in entry.S */
 430                 if (user_stack_base) {
 431                         child_ptregs->r12 = user_stack_base + user_stack_size - 16;
 432                         child_ptregs->ar_bspstore = user_stack_base;
 433                         child_ptregs->ar_rnat = 0;
 434                         child_ptregs->loadrs = 0;
 435                 }
 436         } else {
 437                 /*
 438                  * Note: we simply preserve the relative position of
 439                  * the stack pointer here.  There is no need to
 440                  * allocate a scratch area here, since that will have
 441                  * been taken care of by the caller of sys_clone()
 442                  * already.
 443                  */
 444                 child_ptregs->r12 = (unsigned long) child_ptregs - 16; /* kernel sp */
 445                 child_ptregs->r13 = (unsigned long) p;          /* set `current' pointer */
 446         }
 447         child_stack->ar_bspstore = child_rbs + rbs_size;
 448         if (IS_IA32_PROCESS(regs))
 449                 child_stack->b0 = (unsigned long) &ia32_ret_from_clone;
 450         else
 451                 child_stack->b0 = (unsigned long) &ia64_ret_from_clone;
 452
 453         /* copy parts of thread_struct: */
 454         p->thread.ksp = (unsigned long) child_stack - 16;
 455
 456         /* stop some PSR bits from being inherited.
 457          * the psr.up/psr.pp bits must be cleared on fork but inherited on execve()
 458          * therefore we must specify them explicitly here and not include them in
 459          * IA64_PSR_BITS_TO_CLEAR.
 460          */
 461         child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET)
 462                                  & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP));
 463
 464         /*
 465          * NOTE: The calling convention considers all floating point
 466          * registers in the high partition (fph) to be scratch.  Since
 467          * the only way to get to this point is through a system call,
 468          * we know that the values in fph are all dead.  Hence, there
 469          * is no need to inherit the fph state from the parent to the
 470          * child and all we have to do is to make sure that
 471          * IA64_THREAD_FPH_VALID is cleared in the child.
 472          *
 473          * XXX We could push this optimization a bit further by
 474          * clearing IA64_THREAD_FPH_VALID on ANY system call.
 475          * However, it's not clear this is worth doing.  Also, it
 476          * would be a slight deviation from the normal Linux system
 477          * call behavior where scratch registers are preserved across
 478          * system calls (unless used by the system call itself).
 479          */
 480 #       define THREAD_FLAGS_TO_CLEAR    (IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID \
 481                                          | IA64_THREAD_PM_VALID)
 482 #       define THREAD_FLAGS_TO_SET      0
 483         p->thread.flags = ((current->thread.flags & ~THREAD_FLAGS_TO_CLEAR)
 484                            | THREAD_FLAGS_TO_SET);
 485         ia64_drop_fpu(p);       /* don't pick up stale state from a CPU's fph */
 486 #ifdef CONFIG_IA32_SUPPORT
 487         /*
 488          * If we're cloning an IA32 task then save the IA32 extra
 489          * state from the current task to the new task
 490          */
 491         if (IS_IA32_PROCESS(task_pt_regs(current))) {
 492                 ia32_save_state(p);
 493                 if (clone_flags & CLONE_SETTLS)
 494                         retval = ia32_clone_tls(p, child_ptregs);
 495
 496                 /* Copy partially mapped page list */
 497                 if (!retval)
 498                         retval = ia32_copy_partial_page_list(p, clone_flags);
 499         }
 500 #endif
 501
 502 #ifdef CONFIG_PERFMON
 503         if (current->thread.pfm_context)
 504                 pfm_inherit(p, child_ptregs);
 505 #endif
 506         return retval;
 507 }
 508
 509 static void
 510 do_copy_task_regs (struct task_struct *task, struct unw_frame_info *info, void *arg)
 511 {
 512         unsigned long mask, sp, nat_bits = 0, ip, ar_rnat, urbs_end, cfm;
 513         elf_greg_t *dst = arg;
 514         struct pt_regs *pt;
 515         char nat;
 516         int i;
 517
 518         memset(dst, 0, sizeof(elf_gregset_t));  /* don't leak any kernel bits to user-level */
 519
 520         if (unw_unwind_to_user(info) < 0)
 521                 return;
 522
 523         unw_get_sp(info, &sp);
 524         pt = (struct pt_regs *) (sp + 16);
 525
 526         urbs_end = ia64_get_user_rbs_end(task, pt, &cfm);
 527
 528         if (ia64_sync_user_rbs(task, info->sw, pt->ar_bspstore, urbs_end) < 0)
 529                 return;
 530
 531         ia64_peek(task, info->sw, urbs_end, (long) ia64_rse_rnat_addr((long *) urbs_end),
 532                   &ar_rnat);
 533
 534         /*
 535          * coredump format:
 536          *      r0-r31
 537          *      NaT bits (for r0-r31; bit N == 1 iff rN is a NaT)
 538          *      predicate registers (p0-p63)
 539          *      b0-b7
 540          *      ip cfm user-mask
 541          *      ar.rsc ar.bsp ar.bspstore ar.rnat
 542          *      ar.ccv ar.unat ar.fpsr ar.pfs ar.lc ar.ec
 543          */
 544
 545         /* r0 is zero */
 546         for (i = 1, mask = (1UL << i); i < 32; ++i) {
 547                 unw_get_gr(info, i, &dst[i], &nat);
 548                 if (nat)
 549                         nat_bits |= mask;
 550                 mask <<= 1;
 551         }
 552         dst[32] = nat_bits;
 553         unw_get_pr(info, &dst[33]);
 554
 555         for (i = 0; i < 8; ++i)
 556                 unw_get_br(info, i, &dst[34 + i]);
 557
 558         unw_get_rp(info, &ip);
 559         dst[42] = ip + ia64_psr(pt)->ri;
 560         dst[43] = cfm;
 561         dst[44] = pt->cr_ipsr & IA64_PSR_UM;
 562
 563         unw_get_ar(info, UNW_AR_RSC, &dst[45]);
 564         /*
 565          * For bsp and bspstore, unw_get_ar() would return the kernel
 566          * addresses, but we need the user-level addresses instead:
 567          */
 568         dst[46] = urbs_end;     /* note: by convention PT_AR_BSP points to the end of the urbs! */
 569         dst[47] = pt->ar_bspstore;
 570         dst[48] = ar_rnat;
 571         unw_get_ar(info, UNW_AR_CCV, &dst[49]);
 572         unw_get_ar(info, UNW_AR_UNAT, &dst[50]);
 573         unw_get_ar(info, UNW_AR_FPSR, &dst[51]);
 574         dst[52] = pt->ar_pfs;   /* UNW_AR_PFS is == to pt->cr_ifs for interrupt frames */
 575         unw_get_ar(info, UNW_AR_LC, &dst[53]);
 576         unw_get_ar(info, UNW_AR_EC, &dst[54]);
 577         unw_get_ar(info, UNW_AR_CSD, &dst[55]);
 578         unw_get_ar(info, UNW_AR_SSD, &dst[56]);
 579 }
 580
 581 void
 582 do_dump_task_fpu (struct task_struct *task, struct unw_frame_info *info, void *arg)
 583 {
 584         elf_fpreg_t *dst = arg;
 585         int i;
 586
 587         memset(dst, 0, sizeof(elf_fpregset_t)); /* don't leak any "random" bits */
 588
 589         if (unw_unwind_to_user(info) < 0)
 590                 return;
 591
 592         /* f0 is 0.0, f1 is 1.0 */
 593
 594         for (i = 2; i < 32; ++i)
 595                 unw_get_fr(info, i, dst + i);
 596
 597         ia64_flush_fph(task);
 598         if ((task->thread.flags & IA64_THREAD_FPH_VALID) != 0)
 599                 memcpy(dst + 32, task->thread.fph, 96*16);
 600 }
 601
 602 void
 603 do_copy_regs (struct unw_frame_info *info, void *arg)
 604 {
 605         do_copy_task_regs(current, info, arg);
 606 }
 607
 608 void
 609 do_dump_fpu (struct unw_frame_info *info, void *arg)
 610 {
 611         do_dump_task_fpu(current, info, arg);
 612 }
 613
 614 int
 615 dump_task_regs(struct task_struct *task, elf_gregset_t *regs)
 616 {
 617         struct unw_frame_info tcore_info;
 618
 619         if (current == task) {
 620                 unw_init_running(do_copy_regs, regs);
 621         } else {
 622                 memset(&tcore_info, 0, sizeof(tcore_info));
 623                 unw_init_from_blocked_task(&tcore_info, task);
 624                 do_copy_task_regs(task, &tcore_info, regs);
 625         }
 626         return 1;
 627 }
 628
 629 void
 630 ia64_elf_core_copy_regs (struct pt_regs *pt, elf_gregset_t dst)
 631 {
 632         unw_init_running(do_copy_regs, dst);
 633 }
 634
 635 int
 636 dump_task_fpu (struct task_struct *task, elf_fpregset_t *dst)
 637 {
 638         struct unw_frame_info tcore_info;
 639
 640         if (current == task) {
 641                 unw_init_running(do_dump_fpu, dst);
 642         } else {
 643                 memset(&tcore_info, 0, sizeof(tcore_info));
 644                 unw_init_from_blocked_task(&tcore_info, task);
 645                 do_dump_task_fpu(task, &tcore_info, dst);
 646         }
 647         return 1;
 648 }
 649
 650 int
 651 dump_fpu (struct pt_regs *pt, elf_fpregset_t dst)
 652 {
 653         unw_init_running(do_dump_fpu, dst);
 654         return 1;       /* f0-f31 are always valid so we always return 1 */
 655 }
 656
 657 long
 658 sys_execve (char __user *filename, char __user * __user *argv, char __user * __user *envp,
 659             struct pt_regs *regs)
 660 {
 661         char *fname;
 662         int error;
 663
 664         fname = getname(filename);
 665         error = PTR_ERR(fname);
 666         if (IS_ERR(fname))
 667                 goto out;
 668         error = do_execve(fname, argv, envp, regs);
 669         putname(fname);
 670 out:
 671         return error;
 672 }
 673
 674 pid_t
 675 kernel_thread (int (*fn)(void *), void *arg, unsigned long flags)
 676 {
 677         extern void start_kernel_thread (void);
 678         unsigned long *helper_fptr = (unsigned long *) &start_kernel_thread;
 679         struct {
 680                 struct switch_stack sw;
 681                 struct pt_regs pt;
 682         } regs;
 683
 684         memset(&regs, 0, sizeof(regs));
 685         regs.pt.cr_iip = helper_fptr[0];        /* set entry point (IP) */
 686         regs.pt.r1 = helper_fptr[1];            /* set GP */
 687         regs.pt.r9 = (unsigned long) fn;        /* 1st argument */
 688         regs.pt.r11 = (unsigned long) arg;      /* 2nd argument */
 689         /* Preserve PSR bits, except for bits 32-34 and 37-45, which we can't read.  */
 690         regs.pt.cr_ipsr = ia64_getreg(_IA64_REG_PSR) | IA64_PSR_BN;
 691         regs.pt.cr_ifs = 1UL << 63;             /* mark as valid, empty frame */
 692         regs.sw.ar_fpsr = regs.pt.ar_fpsr = ia64_getreg(_IA64_REG_AR_FPSR);
 693         regs.sw.ar_bspstore = (unsigned long) current + IA64_RBS_OFFSET;
 694         regs.sw.pr = (1 << PRED_KERNEL_STACK);
 695         return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, &regs.pt, 0, NULL, NULL);
 696 }
 697 EXPORT_SYMBOL(kernel_thread);
 698
 699 /* This gets called from kernel_thread() via ia64_invoke_thread_helper().  */
 700 int
 701 kernel_thread_helper (int (*fn)(void *), void *arg)
 702 {
 703 #ifdef CONFIG_IA32_SUPPORT
 704         if (IS_IA32_PROCESS(task_pt_regs(current))) {
 705                 /* A kernel thread is always a 64-bit process. */
 706                 current->thread.map_base  = DEFAULT_MAP_BASE;
 707                 current->thread.task_size = DEFAULT_TASK_SIZE;
 708                 ia64_set_kr(IA64_KR_IO_BASE, current->thread.old_iob);
 709                 ia64_set_kr(IA64_KR_TSSD, current->thread.old_k1);
 710         }
 711 #endif
 712         return (*fn)(arg);
 713 }
 714
 715 /*
 716  * Flush thread state.  This is called when a thread does an execve().
 717  */
 718 void
 719 flush_thread (void)
 720 {
 721         /* drop floating-point and debug-register state if it exists: */
 722         current->thread.flags &= ~(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID);
 723         ia64_drop_fpu(current);
 724 #ifdef CONFIG_IA32_SUPPORT
 725         if (IS_IA32_PROCESS(task_pt_regs(current))) {
 726                 ia32_drop_partial_page_list(current);
 727                 current->thread.task_size = IA32_PAGE_OFFSET;
 728                 set_fs(USER_DS);
 729         }
 730 #endif
 731 }
 732
 733 /*
 734  * Clean up state associated with current thread.  This is called when
 735  * the thread calls exit().
 736  */
 737 void
 738 exit_thread (void)
 739 {
 740
 741         /*
 742          * Remove function-return probe instances associated with this task
 743          * and put them back on the free list. Do not insert an exit probe for
 744          * this function, it will be disabled by kprobe_flush_task if you do.
 745          */
 746         kprobe_flush_task(current);
 747
 748         ia64_drop_fpu(current);
 749 #ifdef CONFIG_PERFMON
 750        /* if needed, stop monitoring and flush state to perfmon context */
 751         if (current->thread.pfm_context)
 752                 pfm_exit_thread(current);
 753
 754         /* free debug register resources */
 755         if (current->thread.flags & IA64_THREAD_DBG_VALID)
 756                 pfm_release_debug_registers(current);
 757 #endif
 758         if (IS_IA32_PROCESS(task_pt_regs(current)))
 759                 ia32_drop_partial_page_list(current);
 760 }
 761
 762 unsigned long
 763 get_wchan (struct task_struct *p)
 764 {
 765         struct unw_frame_info info;
 766         unsigned long ip;
 767         int count = 0;
 768
 769         /*
 770          * Note: p may not be a blocked task (it could be current or
 771          * another process running on some other CPU.  Rather than
 772          * trying to determine if p is really blocked, we just assume
 773          * it's blocked and rely on the unwind routines to fail
 774          * gracefully if the process wasn't really blocked after all.
 775          * --davidm 99/12/15
 776          */
 777         unw_init_from_blocked_task(&info, p);
 778         do {
 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
 812 machine_restart (char *restart_cmd)
 813 {
 814         (void) notify_die(DIE_MACHINE_RESTART, restart_cmd, NULL, 0, 0, 0);
 815         (*efi.reset_system)(EFI_RESET_WARM, 0, 0, NULL);
 816 }
 817
 818 void
 819 machine_halt (void)
 820 {
 821         (void) notify_die(DIE_MACHINE_HALT, "", NULL, 0, 0, 0);
 822         cpu_halt();
 823 }
 824
 825 void
 826 machine_power_off (void)
 827 {
 828         if (pm_power_off)
 829                 pm_power_off();
 830         machine_halt();
 831 }
 832