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