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
  31 #include <asm/cpu.h>
  32 #include <asm/delay.h>
  33 #include <asm/elf.h>
  34 #include <asm/ia32.h>
  35 #include <asm/irq.h>
  36 #include <asm/kexec.h>
  37 #include <asm/pgalloc.h>
  38 #include <asm/processor.h>
  39 #include <asm/sal.h>
  40 #include <asm/tlbflush.h>
  41 #include <asm/uaccess.h>
  42 #include <asm/unwind.h>
  43 #include <asm/user.h>
  44
  45 #include "entry.h"
  46
  47 #ifdef CONFIG_PERFMON
  48 # include <asm/perfmon.h>
  49 #endif
  50
  51 #include "sigframe.h"
  52
  53 void (*ia64_mark_idle)(int);
  54 static DEFINE_PER_CPU(unsigned int, cpu_idle_state);
  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", current->pid, smp_processor_id(), current->comm);
 109         printk("psr : %016lx ifs : %016lx ip  : [<%016lx>]    %s\n",
 110                regs->cr_ipsr, regs->cr_ifs, ip, print_tainted());
 111         print_symbol("ip is at %s\n", ip);
 112         printk("unat: %016lx pfs : %016lx rsc : %016lx\n",
 113                regs->ar_unat, regs->ar_pfs, regs->ar_rsc);
 114         printk("rnat: %016lx bsps: %016lx pr  : %016lx\n",
 115                regs->ar_rnat, regs->ar_bspstore, regs->pr);
 116         printk("ldrs: %016lx ccv : %016lx fpsr: %016lx\n",
 117                regs->loadrs, regs->ar_ccv, regs->ar_fpsr);
 118         printk("csd : %016lx ssd : %016lx\n", regs->ar_csd, regs->ar_ssd);
 119         printk("b0  : %016lx b6  : %016lx b7  : %016lx\n", regs->b0, regs->b6, regs->b7);
 120         printk("f6  : %05lx%016lx f7  : %05lx%016lx\n",
 121                regs->f6.u.bits[1], regs->f6.u.bits[0],
 122                regs->f7.u.bits[1], regs->f7.u.bits[0]);
 123         printk("f8  : %05lx%016lx f9  : %05lx%016lx\n",
 124                regs->f8.u.bits[1], regs->f8.u.bits[0],
 125                regs->f9.u.bits[1], regs->f9.u.bits[0]);
 126         printk("f10 : %05lx%016lx f11 : %05lx%016lx\n",
 127                regs->f10.u.bits[1], regs->f10.u.bits[0],
 128                regs->f11.u.bits[1], regs->f11.u.bits[0]);
 129
 130         printk("r1  : %016lx r2  : %016lx r3  : %016lx\n", regs->r1, regs->r2, regs->r3);
 131         printk("r8  : %016lx r9  : %016lx r10 : %016lx\n", regs->r8, regs->r9, regs->r10);
 132         printk("r11 : %016lx r12 : %016lx r13 : %016lx\n", regs->r11, regs->r12, regs->r13);
 133         printk("r14 : %016lx r15 : %016lx r16 : %016lx\n", regs->r14, regs->r15, regs->r16);
 134         printk("r17 : %016lx r18 : %016lx r19 : %016lx\n", regs->r17, regs->r18, regs->r19);
 135         printk("r20 : %016lx r21 : %016lx r22 : %016lx\n", regs->r20, regs->r21, regs->r22);
 136         printk("r23 : %016lx r24 : %016lx r25 : %016lx\n", regs->r23, regs->r24, regs->r25);
 137         printk("r26 : %016lx r27 : %016lx r28 : %016lx\n", regs->r26, regs->r27, regs->r28);
 138         printk("r29 : %016lx r30 : %016lx r31 : %016lx\n", regs->r29, regs->r30, regs->r31);
 139
 140         if (user_mode(regs)) {
 141                 /* print the stacked registers */
 142                 unsigned long val, *bsp, ndirty;
 143                 int i, sof, is_nat = 0;
 144
 145                 sof = regs->cr_ifs & 0x7f;      /* size of frame */
 146                 ndirty = (regs->loadrs >> 19);
 147                 bsp = ia64_rse_skip_regs((unsigned long *) regs->ar_bspstore, ndirty);
 148                 for (i = 0; i < sof; ++i) {
 149                         get_user(val, (unsigned long __user *) ia64_rse_skip_regs(bsp, i));
 150                         printk("r%-3u:%c%016lx%s", 32 + i, is_nat ? '*' : ' ', val,
 151                                ((i == sof - 1) || (i % 3) == 2) ? "\n" : " ");
 152                 }
 153         } else
 154                 show_stack(NULL, NULL);
 155 }
 156
 157 void
 158 do_notify_resume_user (sigset_t *unused, struct sigscratch *scr, long in_syscall)
 159 {
 160         if (fsys_mode(current, &scr->pt)) {
 161                 /* defer signal-handling etc. until we return to privilege-level 0.  */
 162                 if (!ia64_psr(&scr->pt)->lp)
 163                         ia64_psr(&scr->pt)->lp = 1;
 164                 return;
 165         }
 166
 167 #ifdef CONFIG_PERFMON
 168         if (current->thread.pfm_needs_checking)
 169                 pfm_handle_work();
 170 #endif
 171
 172         /* deal with pending signal delivery */
 173         if (test_thread_flag(TIF_SIGPENDING)||test_thread_flag(TIF_RESTORE_SIGMASK))
 174                 ia64_do_signal(scr, in_syscall);
 175 }
 176
 177 static int pal_halt        = 1;
 178 static int can_do_pal_halt = 1;
 179
 180 static int __init nohalt_setup(char * str)
 181 {
 182         pal_halt = can_do_pal_halt = 0;
 183         return 1;
 184 }
 185 __setup("nohalt", nohalt_setup);
 186
 187 void
 188 update_pal_halt_status(int status)
 189 {
 190         can_do_pal_halt = pal_halt && status;
 191 }
 192
 193 /*
 194  * We use this if we don't have any better idle routine..
 195  */
 196 void
 197 default_idle (void)
 198 {
 199         local_irq_enable();
 200         while (!need_resched()) {
 201                 if (can_do_pal_halt)
 202                         safe_halt();
 203                 else
 204                         cpu_relax();
 205         }
 206 }
 207
 208 #ifdef CONFIG_HOTPLUG_CPU
 209 /* We don't actually take CPU down, just spin without interrupts. */
 210 static inline void play_dead(void)
 211 {
 212         extern void ia64_cpu_local_tick (void);
 213         unsigned int this_cpu = smp_processor_id();
 214
 215         /* Ack it */
 216         __get_cpu_var(cpu_state) = CPU_DEAD;
 217
 218         max_xtp();
 219         local_irq_disable();
 220         idle_task_exit();
 221         ia64_jump_to_sal(&sal_boot_rendez_state[this_cpu]);
 222         /*
 223          * The above is a point of no-return, the processor is
 224          * expected to be in SAL loop now.
 225          */
 226         BUG();
 227 }
 228 #else
 229 static inline void play_dead(void)
 230 {
 231         BUG();
 232 }
 233 #endif /* CONFIG_HOTPLUG_CPU */
 234
 235 void cpu_idle_wait(void)
 236 {
 237         unsigned int cpu, this_cpu = get_cpu();
 238         cpumask_t map;
 239         cpumask_t tmp = current->cpus_allowed;
 240
 241         set_cpus_allowed(current, cpumask_of_cpu(this_cpu));
 242         put_cpu();
 243
 244         cpus_clear(map);
 245         for_each_online_cpu(cpu) {
 246                 per_cpu(cpu_idle_state, cpu) = 1;
 247                 cpu_set(cpu, map);
 248         }
 249
 250         __get_cpu_var(cpu_idle_state) = 0;
 251
 252         wmb();
 253         do {
 254                 ssleep(1);
 255                 for_each_online_cpu(cpu) {
 256                         if (cpu_isset(cpu, map) && !per_cpu(cpu_idle_state, cpu))
 257                                 cpu_clear(cpu, map);
 258                 }
 259                 cpus_and(map, map, cpu_online_map);
 260         } while (!cpus_empty(map));
 261         set_cpus_allowed(current, tmp);
 262 }
 263 EXPORT_SYMBOL_GPL(cpu_idle_wait);
 264
 265 void __attribute__((noreturn))
 266 cpu_idle (void)
 267 {
 268         void (*mark_idle)(int) = ia64_mark_idle;
 269         int cpu = smp_processor_id();
 270
 271         /* endless idle loop with no priority at all */
 272         while (1) {
 273                 if (can_do_pal_halt) {
 274                         current_thread_info()->status &= ~TS_POLLING;
 275                         /*
 276                          * TS_POLLING-cleared state must be visible before we
 277                          * test NEED_RESCHED:
 278                          */
 279                         smp_mb();
 280                 } else {
 281                         current_thread_info()->status |= TS_POLLING;
 282                 }
 283
 284                 if (!need_resched()) {
 285                         void (*idle)(void);
 286 #ifdef CONFIG_SMP
 287                         min_xtp();
 288 #endif
 289                         if (__get_cpu_var(cpu_idle_state))
 290                                 __get_cpu_var(cpu_idle_state) = 0;
 291
 292                         rmb();
 293                         if (mark_idle)
 294                                 (*mark_idle)(1);
 295
 296                         idle = pm_idle;
 297                         if (!idle)
 298                                 idle = default_idle;
 299                         (*idle)();
 300                         if (mark_idle)
 301                                 (*mark_idle)(0);
 302 #ifdef CONFIG_SMP
 303                         normal_xtp();
 304 #endif
 305                 }
 306                 preempt_enable_no_resched();
 307                 schedule();
 308                 preempt_disable();
 309                 check_pgt_cache();
 310                 if (cpu_is_offline(cpu))
 311                         play_dead();
 312         }
 313 }
 314
 315 void
 316 ia64_save_extra (struct task_struct *task)
 317 {
 318 #ifdef CONFIG_PERFMON
 319         unsigned long info;
 320 #endif
 321
 322         if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
 323                 ia64_save_debug_regs(&task->thread.dbr[0]);
 324
 325 #ifdef CONFIG_PERFMON
 326         if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
 327                 pfm_save_regs(task);
 328
 329         info = __get_cpu_var(pfm_syst_info);
 330         if (info & PFM_CPUINFO_SYST_WIDE)
 331                 pfm_syst_wide_update_task(task, info, 0);
 332 #endif
 333
 334 #ifdef CONFIG_IA32_SUPPORT
 335         if (IS_IA32_PROCESS(task_pt_regs(task)))
 336                 ia32_save_state(task);
 337 #endif
 338 }
 339
 340 void
 341 ia64_load_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_load_debug_regs(&task->thread.dbr[0]);
 349
 350 #ifdef CONFIG_PERFMON
 351         if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
 352                 pfm_load_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, 1);
 357 #endif
 358
 359 #ifdef CONFIG_IA32_SUPPORT
 360         if (IS_IA32_PROCESS(task_pt_regs(task)))
 361                 ia32_load_state(task);
 362 #endif
 363 }
 364
 365 /*
 366  * Copy the state of an ia-64 thread.
 367  *
 368  * We get here through the following  call chain:
 369  *
 370  *      from user-level:        from kernel:
 371  *
 372  *      <clone syscall>         <some kernel call frames>
 373  *      sys_clone                  :
 374  *      do_fork                 do_fork
 375  *      copy_thread             copy_thread
 376  *
 377  * This means that the stack layout is as follows:
 378  *
 379  *      +---------------------+ (highest addr)
 380  *      |   struct pt_regs    |
 381  *      +---------------------+
 382  *      | struct switch_stack |
 383  *      +---------------------+
 384  *      |                     |
 385  *      |    memory stack     |
 386  *      |                     | <-- sp (lowest addr)
 387  *      +---------------------+
 388  *
 389  * Observe that we copy the unat values that are in pt_regs and switch_stack.  Spilling an
 390  * integer to address X causes bit N in ar.unat to be set to the NaT bit of the register,
 391  * with N=(X & 0x1ff)/8.  Thus, copying the unat value preserves the NaT bits ONLY if the
 392  * pt_regs structure in the parent is congruent to that of the child, modulo 512.  Since
 393  * the stack is page aligned and the page size is at least 4KB, this is always the case,
 394  * so there is nothing to worry about.
 395  */
 396 int
 397 copy_thread (int nr, unsigned long clone_flags,
 398              unsigned long user_stack_base, unsigned long user_stack_size,
 399              struct task_struct *p, struct pt_regs *regs)
 400 {
 401         extern char ia64_ret_from_clone, ia32_ret_from_clone;
 402         struct switch_stack *child_stack, *stack;
 403         unsigned long rbs, child_rbs, rbs_size;
 404         struct pt_regs *child_ptregs;
 405         int retval = 0;
 406
 407 #ifdef CONFIG_SMP
 408         /*
 409          * For SMP idle threads, fork_by_hand() calls do_fork with
 410          * NULL regs.
 411          */
 412         if (!regs)
 413                 return 0;
 414 #endif
 415
 416         stack = ((struct switch_stack *) regs) - 1;
 417
 418         child_ptregs = (struct pt_regs *) ((unsigned long) p + IA64_STK_OFFSET) - 1;
 419         child_stack = (struct switch_stack *) child_ptregs - 1;
 420
 421         /* copy parent's switch_stack & pt_regs to child: */
 422         memcpy(child_stack, stack, sizeof(*child_ptregs) + sizeof(*child_stack));
 423
 424         rbs = (unsigned long) current + IA64_RBS_OFFSET;
 425         child_rbs = (unsigned long) p + IA64_RBS_OFFSET;
 426         rbs_size = stack->ar_bspstore - rbs;
 427
 428         /* copy the parent's register backing store to the child: */
 429         memcpy((void *) child_rbs, (void *) rbs, rbs_size);
 430
 431         if (likely(user_mode(child_ptregs))) {
 432                 if ((clone_flags & CLONE_SETTLS) && !IS_IA32_PROCESS(regs))
 433                         child_ptregs->r13 = regs->r16;  /* see sys_clone2() in entry.S */
 434                 if (user_stack_base) {
 435                         child_ptregs->r12 = user_stack_base + user_stack_size - 16;
 436                         child_ptregs->ar_bspstore = user_stack_base;
 437                         child_ptregs->ar_rnat = 0;
 438                         child_ptregs->loadrs = 0;
 439                 }
 440         } else {
 441                 /*
 442                  * Note: we simply preserve the relative position of
 443                  * the stack pointer here.  There is no need to
 444                  * allocate a scratch area here, since that will have
 445                  * been taken care of by the caller of sys_clone()
 446                  * already.
 447                  */
 448                 child_ptregs->r12 = (unsigned long) child_ptregs - 16; /* kernel sp */
 449                 child_ptregs->r13 = (unsigned long) p;          /* set `current' pointer */
 450         }
 451         child_stack->ar_bspstore = child_rbs + rbs_size;
 452         if (IS_IA32_PROCESS(regs))
 453                 child_stack->b0 = (unsigned long) &ia32_ret_from_clone;
 454         else
 455                 child_stack->b0 = (unsigned long) &ia64_ret_from_clone;
 456
 457         /* copy parts of thread_struct: */
 458         p->thread.ksp = (unsigned long) child_stack - 16;
 459
 460         /* stop some PSR bits from being inherited.
 461          * the psr.up/psr.pp bits must be cleared on fork but inherited on execve()
 462          * therefore we must specify them explicitly here and not include them in
 463          * IA64_PSR_BITS_TO_CLEAR.
 464          */
 465         child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET)
 466                                  & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP));
 467
 468         /*
 469          * NOTE: The calling convention considers all floating point
 470          * registers in the high partition (fph) to be scratch.  Since
 471          * the only way to get to this point is through a system call,
 472          * we know that the values in fph are all dead.  Hence, there
 473          * is no need to inherit the fph state from the parent to the
 474          * child and all we have to do is to make sure that
 475          * IA64_THREAD_FPH_VALID is cleared in the child.
 476          *
 477          * XXX We could push this optimization a bit further by
 478          * clearing IA64_THREAD_FPH_VALID on ANY system call.
 479          * However, it's not clear this is worth doing.  Also, it
 480          * would be a slight deviation from the normal Linux system
 481          * call behavior where scratch registers are preserved across
 482          * system calls (unless used by the system call itself).
 483          */
 484 #       define THREAD_FLAGS_TO_CLEAR    (IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID \
 485                                          | IA64_THREAD_PM_VALID)
 486 #       define THREAD_FLAGS_TO_SET      0
 487         p->thread.flags = ((current->thread.flags & ~THREAD_FLAGS_TO_CLEAR)
 488                            | THREAD_FLAGS_TO_SET);
 489         ia64_drop_fpu(p);       /* don't pick up stale state from a CPU's fph */
 490 #ifdef CONFIG_IA32_SUPPORT
 491         /*
 492          * If we're cloning an IA32 task then save the IA32 extra
 493          * state from the current task to the new task
 494          */
 495         if (IS_IA32_PROCESS(task_pt_regs(current))) {
 496                 ia32_save_state(p);
 497                 if (clone_flags & CLONE_SETTLS)
 498                         retval = ia32_clone_tls(p, child_ptregs);
 499
 500                 /* Copy partially mapped page list */
 501                 if (!retval)
 502                         retval = ia32_copy_partial_page_list(p, clone_flags);
 503         }
 504 #endif
 505
 506 #ifdef CONFIG_PERFMON
 507         if (current->thread.pfm_context)
 508                 pfm_inherit(p, child_ptregs);
 509 #endif
 510         return retval;
 511 }
 512
 513 static void
 514 do_copy_task_regs (struct task_struct *task, struct unw_frame_info *info, void *arg)
 515 {
 516         unsigned long mask, sp, nat_bits = 0, ip, ar_rnat, urbs_end, cfm;
 517         elf_greg_t *dst = arg;
 518         struct pt_regs *pt;
 519         char nat;
 520         int i;
 521
 522         memset(dst, 0, sizeof(elf_gregset_t));  /* don't leak any kernel bits to user-level */
 523
 524         if (unw_unwind_to_user(info) < 0)
 525                 return;
 526
 527         unw_get_sp(info, &sp);
 528         pt = (struct pt_regs *) (sp + 16);
 529
 530         urbs_end = ia64_get_user_rbs_end(task, pt, &cfm);
 531
 532         if (ia64_sync_user_rbs(task, info->sw, pt->ar_bspstore, urbs_end) < 0)
 533                 return;
 534
 535         ia64_peek(task, info->sw, urbs_end, (long) ia64_rse_rnat_addr((long *) urbs_end),
 536                   &ar_rnat);
 537
 538         /*
 539          * coredump format:
 540          *      r0-r31
 541          *      NaT bits (for r0-r31; bit N == 1 iff rN is a NaT)
 542          *      predicate registers (p0-p63)
 543          *      b0-b7
 544          *      ip cfm user-mask
 545          *      ar.rsc ar.bsp ar.bspstore ar.rnat
 546          *      ar.ccv ar.unat ar.fpsr ar.pfs ar.lc ar.ec
 547          */
 548
 549         /* r0 is zero */
 550         for (i = 1, mask = (1UL << i); i < 32; ++i) {
 551                 unw_get_gr(info, i, &dst[i], &nat);
 552                 if (nat)
 553                         nat_bits |= mask;
 554                 mask <<= 1;
 555         }
 556         dst[32] = nat_bits;
 557         unw_get_pr(info, &dst[33]);
 558
 559         for (i = 0; i < 8; ++i)
 560                 unw_get_br(info, i, &dst[34 + i]);
 561
 562         unw_get_rp(info, &ip);
 563         dst[42] = ip + ia64_psr(pt)->ri;
 564         dst[43] = cfm;
 565         dst[44] = pt->cr_ipsr & IA64_PSR_UM;
 566
 567         unw_get_ar(info, UNW_AR_RSC, &dst[45]);
 568         /*
 569          * For bsp and bspstore, unw_get_ar() would return the kernel
 570          * addresses, but we need the user-level addresses instead:
 571          */
 572         dst[46] = urbs_end;     /* note: by convention PT_AR_BSP points to the end of the urbs! */
 573         dst[47] = pt->ar_bspstore;
 574         dst[48] = ar_rnat;
 575         unw_get_ar(info, UNW_AR_CCV, &dst[49]);
 576         unw_get_ar(info, UNW_AR_UNAT, &dst[50]);
 577         unw_get_ar(info, UNW_AR_FPSR, &dst[51]);
 578         dst[52] = pt->ar_pfs;   /* UNW_AR_PFS is == to pt->cr_ifs for interrupt frames */
 579         unw_get_ar(info, UNW_AR_LC, &dst[53]);
 580         unw_get_ar(info, UNW_AR_EC, &dst[54]);
 581         unw_get_ar(info, UNW_AR_CSD, &dst[55]);
 582         unw_get_ar(info, UNW_AR_SSD, &dst[56]);
 583 }
 584
 585 void
 586 do_dump_task_fpu (struct task_struct *task, struct unw_frame_info *info, void *arg)
 587 {
 588         elf_fpreg_t *dst = arg;
 589         int i;
 590
 591         memset(dst, 0, sizeof(elf_fpregset_t)); /* don't leak any "random" bits */
 592
 593         if (unw_unwind_to_user(info) < 0)
 594                 return;
 595
 596         /* f0 is 0.0, f1 is 1.0 */
 597
 598         for (i = 2; i < 32; ++i)
 599                 unw_get_fr(info, i, dst + i);
 600
 601         ia64_flush_fph(task);
 602         if ((task->thread.flags & IA64_THREAD_FPH_VALID) != 0)
 603                 memcpy(dst + 32, task->thread.fph, 96*16);
 604 }
 605
 606 void
 607 do_copy_regs (struct unw_frame_info *info, void *arg)
 608 {
 609         do_copy_task_regs(current, info, arg);
 610 }
 611
 612 void
 613 do_dump_fpu (struct unw_frame_info *info, void *arg)
 614 {
 615         do_dump_task_fpu(current, info, arg);
 616 }
 617
 618 int
 619 dump_task_regs(struct task_struct *task, elf_gregset_t *regs)
 620 {
 621         struct unw_frame_info tcore_info;
 622
 623         if (current == task) {
 624                 unw_init_running(do_copy_regs, regs);
 625         } else {
 626                 memset(&tcore_info, 0, sizeof(tcore_info));
 627                 unw_init_from_blocked_task(&tcore_info, task);
 628                 do_copy_task_regs(task, &tcore_info, regs);
 629         }
 630         return 1;
 631 }
 632
 633 void
 634 ia64_elf_core_copy_regs (struct pt_regs *pt, elf_gregset_t dst)
 635 {
 636         unw_init_running(do_copy_regs, dst);
 637 }
 638
 639 int
 640 dump_task_fpu (struct task_struct *task, elf_fpregset_t *dst)
 641 {
 642         struct unw_frame_info tcore_info;
 643
 644         if (current == task) {
 645                 unw_init_running(do_dump_fpu, dst);
 646         } else {
 647                 memset(&tcore_info, 0, sizeof(tcore_info));
 648                 unw_init_from_blocked_task(&tcore_info, task);
 649                 do_dump_task_fpu(task, &tcore_info, dst);
 650         }
 651         return 1;
 652 }
 653
 654 int
 655 dump_fpu (struct pt_regs *pt, elf_fpregset_t dst)
 656 {
 657         unw_init_running(do_dump_fpu, dst);
 658         return 1;       /* f0-f31 are always valid so we always return 1 */
 659 }
 660
 661 long
 662 sys_execve (char __user *filename, char __user * __user *argv, char __user * __user *envp,
 663             struct pt_regs *regs)
 664 {
 665         char *fname;
 666         int error;
 667
 668         fname = getname(filename);
 669         error = PTR_ERR(fname);
 670         if (IS_ERR(fname))
 671                 goto out;
 672         error = do_execve(fname, argv, envp, regs);
 673         putname(fname);
 674 out:
 675         return error;
 676 }
 677
 678 pid_t
 679 kernel_thread (int (*fn)(void *), void *arg, unsigned long flags)
 680 {
 681         extern void start_kernel_thread (void);
 682         unsigned long *helper_fptr = (unsigned long *) &start_kernel_thread;
 683         struct {
 684                 struct switch_stack sw;
 685                 struct pt_regs pt;
 686         } regs;
 687
 688         memset(&regs, 0, sizeof(regs));
 689         regs.pt.cr_iip = helper_fptr[0];        /* set entry point (IP) */
 690         regs.pt.r1 = helper_fptr[1];            /* set GP */
 691         regs.pt.r9 = (unsigned long) fn;        /* 1st argument */
 692         regs.pt.r11 = (unsigned long) arg;      /* 2nd argument */
 693         /* Preserve PSR bits, except for bits 32-34 and 37-45, which we can't read.  */
 694         regs.pt.cr_ipsr = ia64_getreg(_IA64_REG_PSR) | IA64_PSR_BN;
 695         regs.pt.cr_ifs = 1UL << 63;             /* mark as valid, empty frame */
 696         regs.sw.ar_fpsr = regs.pt.ar_fpsr = ia64_getreg(_IA64_REG_AR_FPSR);
 697         regs.sw.ar_bspstore = (unsigned long) current + IA64_RBS_OFFSET;
 698         regs.sw.pr = (1 << PRED_KERNEL_STACK);
 699         return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, &regs.pt, 0, NULL, NULL);
 700 }
 701 EXPORT_SYMBOL(kernel_thread);
 702
 703 /* This gets called from kernel_thread() via ia64_invoke_thread_helper().  */
 704 int
 705 kernel_thread_helper (int (*fn)(void *), void *arg)
 706 {
 707 #ifdef CONFIG_IA32_SUPPORT
 708         if (IS_IA32_PROCESS(task_pt_regs(current))) {
 709                 /* A kernel thread is always a 64-bit process. */
 710                 current->thread.map_base  = DEFAULT_MAP_BASE;
 711                 current->thread.task_size = DEFAULT_TASK_SIZE;
 712                 ia64_set_kr(IA64_KR_IO_BASE, current->thread.old_iob);
 713                 ia64_set_kr(IA64_KR_TSSD, current->thread.old_k1);
 714         }
 715 #endif
 716         return (*fn)(arg);
 717 }
 718
 719 /*
 720  * Flush thread state.  This is called when a thread does an execve().
 721  */
 722 void
 723 flush_thread (void)
 724 {
 725         /* drop floating-point and debug-register state if it exists: */
 726         current->thread.flags &= ~(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID);
 727         ia64_drop_fpu(current);
 728 #ifdef CONFIG_IA32_SUPPORT
 729         if (IS_IA32_PROCESS(task_pt_regs(current))) {
 730                 ia32_drop_partial_page_list(current);
 731                 current->thread.task_size = IA32_PAGE_OFFSET;
 732                 set_fs(USER_DS);
 733         }
 734 #endif
 735 }
 736
 737 /*
 738  * Clean up state associated with current thread.  This is called when
 739  * the thread calls exit().
 740  */
 741 void
 742 exit_thread (void)
 743 {
 744
 745         ia64_drop_fpu(current);
 746 #ifdef CONFIG_PERFMON
 747        /* if needed, stop monitoring and flush state to perfmon context */
 748         if (current->thread.pfm_context)
 749                 pfm_exit_thread(current);
 750
 751         /* free debug register resources */
 752         if (current->thread.flags & IA64_THREAD_DBG_VALID)
 753                 pfm_release_debug_registers(current);
 754 #endif
 755         if (IS_IA32_PROCESS(task_pt_regs(current)))
 756                 ia32_drop_partial_page_list(current);
 757 }
 758
 759 unsigned long
 760 get_wchan (struct task_struct *p)
 761 {
 762         struct unw_frame_info info;
 763         unsigned long ip;
 764         int count = 0;
 765
 766         if (!p || p == current || p->state == TASK_RUNNING)
 767                 return 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 (p->state == TASK_RUNNING)
 780                         return 0;
 781                 if (unw_unwind(&info) < 0)
 782                         return 0;
 783                 unw_get_ip(&info, &ip);
 784                 if (!in_sched_functions(ip))
 785                         return ip;
 786         } while (count++ < 16);
 787         return 0;
 788 }
 789
 790 void
 791 cpu_halt (void)
 792 {
 793         pal_power_mgmt_info_u_t power_info[8];
 794         unsigned long min_power;
 795         int i, min_power_state;
 796
 797         if (ia64_pal_halt_info(power_info) != 0)
 798                 return;
 799
 800         min_power_state = 0;
 801         min_power = power_info[0].pal_power_mgmt_info_s.power_consumption;
 802         for (i = 1; i < 8; ++i)
 803                 if (power_info[i].pal_power_mgmt_info_s.im
 804                     && power_info[i].pal_power_mgmt_info_s.power_consumption < min_power) {
 805                         min_power = power_info[i].pal_power_mgmt_info_s.power_consumption;
 806                         min_power_state = i;
 807                 }
 808
 809         while (1)
 810                 ia64_pal_halt(min_power_state);
 811 }
 812
 813 void machine_shutdown(void)
 814 {
 815 #ifdef CONFIG_HOTPLUG_CPU
 816         int cpu;
 817
 818         for_each_online_cpu(cpu) {
 819                 if (cpu != smp_processor_id())
 820                         cpu_down(cpu);
 821         }
 822 #endif
 823 #ifdef CONFIG_KEXEC
 824         kexec_disable_iosapic();
 825 #endif
 826 }
 827
 828 void
 829 machine_restart (char *restart_cmd)
 830 {
 831         (void) notify_die(DIE_MACHINE_RESTART, restart_cmd, NULL, 0, 0, 0);
 832         (*efi.reset_system)(EFI_RESET_WARM, 0, 0, NULL);
 833 }
 834
 835 void
 836 machine_halt (void)
 837 {
 838         (void) notify_die(DIE_MACHINE_HALT, "", NULL, 0, 0, 0);
 839         cpu_halt();
 840 }
 841
 842 void
 843 machine_power_off (void)
 844 {
 845         if (pm_power_off)
 846                 pm_power_off();
 847         machine_halt();
 848 }
 849