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
  13 #include <linux/cpu.h>
  14 #include <linux/pm.h>
  15 #include <linux/elf.h>
  16 #include <linux/errno.h>
  17 #include <linux/kallsyms.h>
  18 #include <linux/kernel.h>
  19 #include <linux/mm.h>
  20 #include <linux/module.h>
  21 #include <linux/notifier.h>
  22 #include <linux/personality.h>
  23 #include <linux/sched.h>
  24 #include <linux/slab.h>
  25 #include <linux/smp_lock.h>
  26 #include <linux/stddef.h>
  27 #include <linux/thread_info.h>
  28 #include <linux/unistd.h>
  29 #include <linux/efi.h>
  30 #include <linux/interrupt.h>
  31 #include <linux/delay.h>
  32
  33 #include <asm/cpu.h>
  34 #include <asm/delay.h>
  35 #include <asm/elf.h>
  36 #include <asm/ia32.h>
  37 #include <asm/irq.h>
  38 #include <asm/kdebug.h>
  39 #include <asm/pgalloc.h>
  40 #include <asm/processor.h>
  41 #include <asm/sal.h>
  42 #include <asm/tlbflush.h>
  43 #include <asm/uaccess.h>
  44 #include <asm/unwind.h>
  45 #include <asm/user.h>
  46
  47 #include "entry.h"
  48
  49 #ifdef CONFIG_PERFMON
  50 # include <asm/perfmon.h>
  51 #endif
  52
  53 #include "sigframe.h"
  54
  55 void (*ia64_mark_idle)(int);
  56 static DEFINE_PER_CPU(unsigned int, cpu_idle_state);
  57
  58 unsigned long boot_option_idle_override = 0;
  59 EXPORT_SYMBOL(boot_option_idle_override);
  60
  61 void
  62 ia64_do_show_stack (struct unw_frame_info *info, void *arg)
  63 {
  64         unsigned long ip, sp, bsp;
  65         char buf[128];                  /* don't make it so big that it overflows the stack! */
  66
  67         printk("\nCall Trace:\n");
  68         do {
  69                 unw_get_ip(info, &ip);
  70                 if (ip == 0)
  71                         break;
  72
  73                 unw_get_sp(info, &sp);
  74                 unw_get_bsp(info, &bsp);
  75                 snprintf(buf, sizeof(buf),
  76                          " [<%016lx>] %%s\n"
  77                          "                                sp=%016lx bsp=%016lx\n",
  78                          ip, sp, bsp);
  79                 print_symbol(buf, ip);
  80         } while (unw_unwind(info) >= 0);
  81 }
  82
  83 void
  84 show_stack (struct task_struct *task, unsigned long *sp)
  85 {
  86         if (!task)
  87                 unw_init_running(ia64_do_show_stack, NULL);
  88         else {
  89                 struct unw_frame_info info;
  90
  91                 unw_init_from_blocked_task(&info, task);
  92                 ia64_do_show_stack(&info, NULL);
  93         }
  94 }
  95
  96 void
  97 dump_stack (void)
  98 {
  99         show_stack(NULL, NULL);
 100 }
 101
 102 EXPORT_SYMBOL(dump_stack);
 103
 104 void
 105 show_regs (struct pt_regs *regs)
 106 {
 107         unsigned long ip = regs->cr_iip + ia64_psr(regs)->ri;
 108
 109         print_modules();
 110         printk("\nPid: %d, CPU %d, comm: %20s\n", current->pid, smp_processor_id(), current->comm);
 111         printk("psr : %016lx ifs : %016lx ip  : [<%016lx>]    %s\n",
 112                regs->cr_ipsr, regs->cr_ifs, ip, print_tainted());
 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
 160 do_notify_resume_user (sigset_t *oldset, struct sigscratch *scr, long in_syscall)
 161 {
 162         if (fsys_mode(current, &scr->pt)) {
 163                 /* defer signal-handling etc. until we return to privilege-level 0.  */
 164                 if (!ia64_psr(&scr->pt)->lp)
 165                         ia64_psr(&scr->pt)->lp = 1;
 166                 return;
 167         }
 168
 169 #ifdef CONFIG_PERFMON
 170         if (current->thread.pfm_needs_checking)
 171                 pfm_handle_work();
 172 #endif
 173
 174         /* deal with pending signal delivery */
 175         if (test_thread_flag(TIF_SIGPENDING))
 176                 ia64_do_signal(oldset, scr, in_syscall);
 177 }
 178
 179 static int pal_halt        = 1;
 180 static int can_do_pal_halt = 1;
 181
 182 static int __init nohalt_setup(char * str)
 183 {
 184         pal_halt = can_do_pal_halt = 0;
 185         return 1;
 186 }
 187 __setup("nohalt", nohalt_setup);
 188
 189 void
 190 update_pal_halt_status(int status)
 191 {
 192         can_do_pal_halt = pal_halt && status;
 193 }
 194
 195 /*
 196  * We use this if we don't have any better idle routine..
 197  */
 198 void
 199 default_idle (void)
 200 {
 201         local_irq_enable();
 202         while (!need_resched()) {
 203                 if (can_do_pal_halt)
 204                         safe_halt();
 205                 else
 206                         cpu_relax();
 207         }
 208 }
 209
 210 #ifdef CONFIG_HOTPLUG_CPU
 211 /* We don't actually take CPU down, just spin without interrupts. */
 212 static inline void play_dead(void)
 213 {
 214         extern void ia64_cpu_local_tick (void);
 215         unsigned int this_cpu = smp_processor_id();
 216
 217         /* Ack it */
 218         __get_cpu_var(cpu_state) = CPU_DEAD;
 219
 220         max_xtp();
 221         local_irq_disable();
 222         idle_task_exit();
 223         ia64_jump_to_sal(&sal_boot_rendez_state[this_cpu]);
 224         /*
 225          * The above is a point of no-return, the processor is
 226          * expected to be in SAL loop now.
 227          */
 228         BUG();
 229 }
 230 #else
 231 static inline void play_dead(void)
 232 {
 233         BUG();
 234 }
 235 #endif /* CONFIG_HOTPLUG_CPU */
 236
 237 void cpu_idle_wait(void)
 238 {
 239         unsigned int cpu, this_cpu = get_cpu();
 240         cpumask_t map;
 241
 242         set_cpus_allowed(current, cpumask_of_cpu(this_cpu));
 243         put_cpu();
 244
 245         cpus_clear(map);
 246         for_each_online_cpu(cpu) {
 247                 per_cpu(cpu_idle_state, cpu) = 1;
 248                 cpu_set(cpu, map);
 249         }
 250
 251         __get_cpu_var(cpu_idle_state) = 0;
 252
 253         wmb();
 254         do {
 255                 ssleep(1);
 256                 for_each_online_cpu(cpu) {
 257                         if (cpu_isset(cpu, map) && !per_cpu(cpu_idle_state, cpu))
 258                                 cpu_clear(cpu, map);
 259                 }
 260                 cpus_and(map, map, cpu_online_map);
 261         } while (!cpus_empty(map));
 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                 else
 276                         current_thread_info()->status |= TS_POLLING;
 277
 278                 if (!need_resched()) {
 279                         void (*idle)(void);
 280 #ifdef CONFIG_SMP
 281                         min_xtp();
 282 #endif
 283                         if (__get_cpu_var(cpu_idle_state))
 284                                 __get_cpu_var(cpu_idle_state) = 0;
 285
 286                         rmb();
 287                         if (mark_idle)
 288                                 (*mark_idle)(1);
 289
 290                         idle = pm_idle;
 291                         if (!idle)
 292                                 idle = default_idle;
 293                         (*idle)();
 294                         if (mark_idle)
 295                                 (*mark_idle)(0);
 296 #ifdef CONFIG_SMP
 297                         normal_xtp();
 298 #endif
 299                 }
 300                 preempt_enable_no_resched();
 301                 schedule();
 302                 preempt_disable();
 303                 check_pgt_cache();
 304                 if (cpu_is_offline(cpu))
 305                         play_dead();
 306         }
 307 }
 308
 309 void
 310 ia64_save_extra (struct task_struct *task)
 311 {
 312 #ifdef CONFIG_PERFMON
 313         unsigned long info;
 314 #endif
 315
 316         if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
 317                 ia64_save_debug_regs(&task->thread.dbr[0]);
 318
 319 #ifdef CONFIG_PERFMON
 320         if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
 321                 pfm_save_regs(task);
 322
 323         info = __get_cpu_var(pfm_syst_info);
 324         if (info & PFM_CPUINFO_SYST_WIDE)
 325                 pfm_syst_wide_update_task(task, info, 0);
 326 #endif
 327
 328 #ifdef CONFIG_IA32_SUPPORT
 329         if (IS_IA32_PROCESS(task_pt_regs(task)))
 330                 ia32_save_state(task);
 331 #endif
 332 }
 333
 334 void
 335 ia64_load_extra (struct task_struct *task)
 336 {
 337 #ifdef CONFIG_PERFMON
 338         unsigned long info;
 339 #endif
 340
 341         if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
 342                 ia64_load_debug_regs(&task->thread.dbr[0]);
 343
 344 #ifdef CONFIG_PERFMON
 345         if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
 346                 pfm_load_regs(task);
 347
 348         info = __get_cpu_var(pfm_syst_info);
 349         if (info & PFM_CPUINFO_SYST_WIDE)
 350                 pfm_syst_wide_update_task(task, info, 1);
 351 #endif
 352
 353 #ifdef CONFIG_IA32_SUPPORT
 354         if (IS_IA32_PROCESS(task_pt_regs(task)))
 355                 ia32_load_state(task);
 356 #endif
 357 }
 358
 359 /*
 360  * Copy the state of an ia-64 thread.
 361  *
 362  * We get here through the following  call chain:
 363  *
 364  *      from user-level:        from kernel:
 365  *
 366  *      <clone syscall>         <some kernel call frames>
 367  *      sys_clone                  :
 368  *      do_fork                 do_fork
 369  *      copy_thread             copy_thread
 370  *
 371  * This means that the stack layout is as follows:
 372  *
 373  *      +---------------------+ (highest addr)
 374  *      |   struct pt_regs    |
 375  *      +---------------------+
 376  *      | struct switch_stack |
 377  *      +---------------------+
 378  *      |                     |
 379  *      |    memory stack     |
 380  *      |                     | <-- sp (lowest addr)
 381  *      +---------------------+
 382  *
 383  * Observe that we copy the unat values that are in pt_regs and switch_stack.  Spilling an
 384  * integer to address X causes bit N in ar.unat to be set to the NaT bit of the register,
 385  * with N=(X & 0x1ff)/8.  Thus, copying the unat value preserves the NaT bits ONLY if the
 386  * pt_regs structure in the parent is congruent to that of the child, modulo 512.  Since
 387  * the stack is page aligned and the page size is at least 4KB, this is always the case,
 388  * so there is nothing to worry about.
 389  */
 390 int
 391 copy_thread (int nr, unsigned long clone_flags,
 392              unsigned long user_stack_base, unsigned long user_stack_size,
 393              struct task_struct *p, struct pt_regs *regs)
 394 {
 395         extern char ia64_ret_from_clone, ia32_ret_from_clone;
 396         struct switch_stack *child_stack, *stack;
 397         unsigned long rbs, child_rbs, rbs_size;
 398         struct pt_regs *child_ptregs;
 399         int retval = 0;
 400
 401 #ifdef CONFIG_SMP
 402         /*
 403          * For SMP idle threads, fork_by_hand() calls do_fork with
 404          * NULL regs.
 405          */
 406         if (!regs)
 407                 return 0;
 408 #endif
 409
 410         stack = ((struct switch_stack *) regs) - 1;
 411
 412         child_ptregs = (struct pt_regs *) ((unsigned long) p + IA64_STK_OFFSET) - 1;
 413         child_stack = (struct switch_stack *) child_ptregs - 1;
 414
 415         /* copy parent's switch_stack & pt_regs to child: */
 416         memcpy(child_stack, stack, sizeof(*child_ptregs) + sizeof(*child_stack));
 417
 418         rbs = (unsigned long) current + IA64_RBS_OFFSET;
 419         child_rbs = (unsigned long) p + IA64_RBS_OFFSET;
 420         rbs_size = stack->ar_bspstore - rbs;
 421
 422         /* copy the parent's register backing store to the child: */
 423         memcpy((void *) child_rbs, (void *) rbs, rbs_size);
 424
 425         if (likely(user_mode(child_ptregs))) {
 426                 if ((clone_flags & CLONE_SETTLS) && !IS_IA32_PROCESS(regs))
 427                         child_ptregs->r13 = regs->r16;  /* see sys_clone2() in entry.S */
 428                 if (user_stack_base) {
 429                         child_ptregs->r12 = user_stack_base + user_stack_size - 16;
 430                         child_ptregs->ar_bspstore = user_stack_base;
 431                         child_ptregs->ar_rnat = 0;
 432                         child_ptregs->loadrs = 0;
 433                 }
 434         } else {
 435                 /*
 436                  * Note: we simply preserve the relative position of
 437                  * the stack pointer here.  There is no need to
 438                  * allocate a scratch area here, since that will have
 439                  * been taken care of by the caller of sys_clone()
 440                  * already.
 441                  */
 442                 child_ptregs->r12 = (unsigned long) child_ptregs - 16; /* kernel sp */
 443                 child_ptregs->r13 = (unsigned long) p;          /* set `current' pointer */
 444         }
 445         child_stack->ar_bspstore = child_rbs + rbs_size;
 446         if (IS_IA32_PROCESS(regs))
 447                 child_stack->b0 = (unsigned long) &ia32_ret_from_clone;
 448         else
 449                 child_stack->b0 = (unsigned long) &ia64_ret_from_clone;
 450
 451         /* copy parts of thread_struct: */
 452         p->thread.ksp = (unsigned long) child_stack - 16;
 453
 454         /* stop some PSR bits from being inherited.
 455          * the psr.up/psr.pp bits must be cleared on fork but inherited on execve()
 456          * therefore we must specify them explicitly here and not include them in
 457          * IA64_PSR_BITS_TO_CLEAR.
 458          */
 459         child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET)
 460                                  & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP));
 461
 462         /*
 463          * NOTE: The calling convention considers all floating point
 464          * registers in the high partition (fph) to be scratch.  Since
 465          * the only way to get to this point is through a system call,
 466          * we know that the values in fph are all dead.  Hence, there
 467          * is no need to inherit the fph state from the parent to the
 468          * child and all we have to do is to make sure that
 469          * IA64_THREAD_FPH_VALID is cleared in the child.
 470          *
 471          * XXX We could push this optimization a bit further by
 472          * clearing IA64_THREAD_FPH_VALID on ANY system call.
 473          * However, it's not clear this is worth doing.  Also, it
 474          * would be a slight deviation from the normal Linux system
 475          * call behavior where scratch registers are preserved across
 476          * system calls (unless used by the system call itself).
 477          */
 478 #       define THREAD_FLAGS_TO_CLEAR    (IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID \
 479                                          | IA64_THREAD_PM_VALID)
 480 #       define THREAD_FLAGS_TO_SET      0
 481         p->thread.flags = ((current->thread.flags & ~THREAD_FLAGS_TO_CLEAR)
 482                            | THREAD_FLAGS_TO_SET);
 483         ia64_drop_fpu(p);       /* don't pick up stale state from a CPU's fph */
 484 #ifdef CONFIG_IA32_SUPPORT
 485         /*
 486          * If we're cloning an IA32 task then save the IA32 extra
 487          * state from the current task to the new task
 488          */
 489         if (IS_IA32_PROCESS(task_pt_regs(current))) {
 490                 ia32_save_state(p);
 491                 if (clone_flags & CLONE_SETTLS)
 492                         retval = ia32_clone_tls(p, child_ptregs);
 493
 494                 /* Copy partially mapped page list */
 495                 if (!retval)
 496                         retval = ia32_copy_partial_page_list(p, clone_flags);
 497         }
 498 #endif
 499
 500 #ifdef CONFIG_PERFMON
 501         if (current->thread.pfm_context)
 502                 pfm_inherit(p, child_ptregs);
 503 #endif
 504         return retval;
 505 }
 506
 507 static void
 508 do_copy_task_regs (struct task_struct *task, struct unw_frame_info *info, void *arg)
 509 {
 510         unsigned long mask, sp, nat_bits = 0, ip, ar_rnat, urbs_end, cfm;
 511         elf_greg_t *dst = arg;
 512         struct pt_regs *pt;
 513         char nat;
 514         int i;
 515
 516         memset(dst, 0, sizeof(elf_gregset_t));  /* don't leak any kernel bits to user-level */
 517
 518         if (unw_unwind_to_user(info) < 0)
 519                 return;
 520
 521         unw_get_sp(info, &sp);
 522         pt = (struct pt_regs *) (sp + 16);
 523
 524         urbs_end = ia64_get_user_rbs_end(task, pt, &cfm);
 525
 526         if (ia64_sync_user_rbs(task, info->sw, pt->ar_bspstore, urbs_end) < 0)
 527                 return;
 528
 529         ia64_peek(task, info->sw, urbs_end, (long) ia64_rse_rnat_addr((long *) urbs_end),
 530                   &ar_rnat);
 531
 532         /*
 533          * coredump format:
 534          *      r0-r31
 535          *      NaT bits (for r0-r31; bit N == 1 iff rN is a NaT)
 536          *      predicate registers (p0-p63)
 537          *      b0-b7
 538          *      ip cfm user-mask
 539          *      ar.rsc ar.bsp ar.bspstore ar.rnat
 540          *      ar.ccv ar.unat ar.fpsr ar.pfs ar.lc ar.ec
 541          */
 542
 543         /* r0 is zero */
 544         for (i = 1, mask = (1UL << i); i < 32; ++i) {
 545                 unw_get_gr(info, i, &dst[i], &nat);
 546                 if (nat)
 547                         nat_bits |= mask;
 548                 mask <<= 1;
 549         }
 550         dst[32] = nat_bits;
 551         unw_get_pr(info, &dst[33]);
 552
 553         for (i = 0; i < 8; ++i)
 554                 unw_get_br(info, i, &dst[34 + i]);
 555
 556         unw_get_rp(info, &ip);
 557         dst[42] = ip + ia64_psr(pt)->ri;
 558         dst[43] = cfm;
 559         dst[44] = pt->cr_ipsr & IA64_PSR_UM;
 560
 561         unw_get_ar(info, UNW_AR_RSC, &dst[45]);
 562         /*
 563          * For bsp and bspstore, unw_get_ar() would return the kernel
 564          * addresses, but we need the user-level addresses instead:
 565          */
 566         dst[46] = urbs_end;     /* note: by convention PT_AR_BSP points to the end of the urbs! */
 567         dst[47] = pt->ar_bspstore;
 568         dst[48] = ar_rnat;
 569         unw_get_ar(info, UNW_AR_CCV, &dst[49]);
 570         unw_get_ar(info, UNW_AR_UNAT, &dst[50]);
 571         unw_get_ar(info, UNW_AR_FPSR, &dst[51]);
 572         dst[52] = pt->ar_pfs;   /* UNW_AR_PFS is == to pt->cr_ifs for interrupt frames */
 573         unw_get_ar(info, UNW_AR_LC, &dst[53]);
 574         unw_get_ar(info, UNW_AR_EC, &dst[54]);
 575         unw_get_ar(info, UNW_AR_CSD, &dst[55]);
 576         unw_get_ar(info, UNW_AR_SSD, &dst[56]);
 577 }
 578
 579 void
 580 do_dump_task_fpu (struct task_struct *task, struct unw_frame_info *info, void *arg)
 581 {
 582         elf_fpreg_t *dst = arg;
 583         int i;
 584
 585         memset(dst, 0, sizeof(elf_fpregset_t)); /* don't leak any "random" bits */
 586
 587         if (unw_unwind_to_user(info) < 0)
 588                 return;
 589
 590         /* f0 is 0.0, f1 is 1.0 */
 591
 592         for (i = 2; i < 32; ++i)
 593                 unw_get_fr(info, i, dst + i);
 594
 595         ia64_flush_fph(task);
 596         if ((task->thread.flags & IA64_THREAD_FPH_VALID) != 0)
 597                 memcpy(dst + 32, task->thread.fph, 96*16);
 598 }
 599
 600 void
 601 do_copy_regs (struct unw_frame_info *info, void *arg)
 602 {
 603         do_copy_task_regs(current, info, arg);
 604 }
 605
 606 void
 607 do_dump_fpu (struct unw_frame_info *info, void *arg)
 608 {
 609         do_dump_task_fpu(current, info, arg);
 610 }
 611
 612 int
 613 dump_task_regs(struct task_struct *task, elf_gregset_t *regs)
 614 {
 615         struct unw_frame_info tcore_info;
 616
 617         if (current == task) {
 618                 unw_init_running(do_copy_regs, regs);
 619         } else {
 620                 memset(&tcore_info, 0, sizeof(tcore_info));
 621                 unw_init_from_blocked_task(&tcore_info, task);
 622                 do_copy_task_regs(task, &tcore_info, regs);
 623         }
 624         return 1;
 625 }
 626
 627 void
 628 ia64_elf_core_copy_regs (struct pt_regs *pt, elf_gregset_t dst)
 629 {
 630         unw_init_running(do_copy_regs, dst);
 631 }
 632
 633 int
 634 dump_task_fpu (struct task_struct *task, elf_fpregset_t *dst)
 635 {
 636         struct unw_frame_info tcore_info;
 637
 638         if (current == task) {
 639                 unw_init_running(do_dump_fpu, dst);
 640         } else {
 641                 memset(&tcore_info, 0, sizeof(tcore_info));
 642                 unw_init_from_blocked_task(&tcore_info, task);
 643                 do_dump_task_fpu(task, &tcore_info, dst);
 644         }
 645         return 1;
 646 }
 647
 648 int
 649 dump_fpu (struct pt_regs *pt, elf_fpregset_t dst)
 650 {
 651         unw_init_running(do_dump_fpu, dst);
 652         return 1;       /* f0-f31 are always valid so we always return 1 */
 653 }
 654
 655 long
 656 sys_execve (char __user *filename, char __user * __user *argv, char __user * __user *envp,
 657             struct pt_regs *regs)
 658 {
 659         char *fname;
 660         int error;
 661
 662         fname = getname(filename);
 663         error = PTR_ERR(fname);
 664         if (IS_ERR(fname))
 665                 goto out;
 666         error = do_execve(fname, argv, envp, regs);
 667         putname(fname);
 668 out:
 669         return error;
 670 }
 671
 672 pid_t
 673 kernel_thread (int (*fn)(void *), void *arg, unsigned long flags)
 674 {
 675         extern void start_kernel_thread (void);
 676         unsigned long *helper_fptr = (unsigned long *) &start_kernel_thread;
 677         struct {
 678                 struct switch_stack sw;
 679                 struct pt_regs pt;
 680         } regs;
 681
 682         memset(&regs, 0, sizeof(regs));
 683         regs.pt.cr_iip = helper_fptr[0];        /* set entry point (IP) */
 684         regs.pt.r1 = helper_fptr[1];            /* set GP */
 685         regs.pt.r9 = (unsigned long) fn;        /* 1st argument */
 686         regs.pt.r11 = (unsigned long) arg;      /* 2nd argument */
 687         /* Preserve PSR bits, except for bits 32-34 and 37-45, which we can't read.  */
 688         regs.pt.cr_ipsr = ia64_getreg(_IA64_REG_PSR) | IA64_PSR_BN;
 689         regs.pt.cr_ifs = 1UL << 63;             /* mark as valid, empty frame */
 690         regs.sw.ar_fpsr = regs.pt.ar_fpsr = ia64_getreg(_IA64_REG_AR_FPSR);
 691         regs.sw.ar_bspstore = (unsigned long) current + IA64_RBS_OFFSET;
 692         regs.sw.pr = (1 << PRED_KERNEL_STACK);
 693         return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, &regs.pt, 0, NULL, NULL);
 694 }
 695 EXPORT_SYMBOL(kernel_thread);
 696
 697 /* This gets called from kernel_thread() via ia64_invoke_thread_helper().  */
 698 int
 699 kernel_thread_helper (int (*fn)(void *), void *arg)
 700 {
 701 #ifdef CONFIG_IA32_SUPPORT
 702         if (IS_IA32_PROCESS(task_pt_regs(current))) {
 703                 /* A kernel thread is always a 64-bit process. */
 704                 current->thread.map_base  = DEFAULT_MAP_BASE;
 705                 current->thread.task_size = DEFAULT_TASK_SIZE;
 706                 ia64_set_kr(IA64_KR_IO_BASE, current->thread.old_iob);
 707                 ia64_set_kr(IA64_KR_TSSD, current->thread.old_k1);
 708         }
 709 #endif
 710         return (*fn)(arg);
 711 }
 712
 713 /*
 714  * Flush thread state.  This is called when a thread does an execve().
 715  */
 716 void
 717 flush_thread (void)
 718 {
 719         /* drop floating-point and debug-register state if it exists: */
 720         current->thread.flags &= ~(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID);
 721         ia64_drop_fpu(current);
 722 #ifdef CONFIG_IA32_SUPPORT
 723         if (IS_IA32_PROCESS(task_pt_regs(current))) {
 724                 ia32_drop_partial_page_list(current);
 725                 current->thread.task_size = IA32_PAGE_OFFSET;
 726                 set_fs(USER_DS);
 727         }
 728 #endif
 729 }
 730
 731 /*
 732  * Clean up state associated with current thread.  This is called when
 733  * the thread calls exit().
 734  */
 735 void
 736 exit_thread (void)
 737 {
 738
 739         ia64_drop_fpu(current);
 740 #ifdef CONFIG_PERFMON
 741        /* if needed, stop monitoring and flush state to perfmon context */
 742         if (current->thread.pfm_context)
 743                 pfm_exit_thread(current);
 744
 745         /* free debug register resources */
 746         if (current->thread.flags & IA64_THREAD_DBG_VALID)
 747                 pfm_release_debug_registers(current);
 748 #endif
 749         if (IS_IA32_PROCESS(task_pt_regs(current)))
 750                 ia32_drop_partial_page_list(current);
 751 }
 752
 753 unsigned long
 754 get_wchan (struct task_struct *p)
 755 {
 756         struct unw_frame_info info;
 757         unsigned long ip;
 758         int count = 0;
 759
 760         /*
 761          * Note: p may not be a blocked task (it could be current or
 762          * another process running on some other CPU.  Rather than
 763          * trying to determine if p is really blocked, we just assume
 764          * it's blocked and rely on the unwind routines to fail
 765          * gracefully if the process wasn't really blocked after all.
 766          * --davidm 99/12/15
 767          */
 768         unw_init_from_blocked_task(&info, p);
 769         do {
 770                 if (unw_unwind(&info) < 0)
 771                         return 0;
 772                 unw_get_ip(&info, &ip);
 773                 if (!in_sched_functions(ip))
 774                         return ip;
 775         } while (count++ < 16);
 776         return 0;
 777 }
 778
 779 void
 780 cpu_halt (void)
 781 {
 782         pal_power_mgmt_info_u_t power_info[8];
 783         unsigned long min_power;
 784         int i, min_power_state;
 785
 786         if (ia64_pal_halt_info(power_info) != 0)
 787                 return;
 788
 789         min_power_state = 0;
 790         min_power = power_info[0].pal_power_mgmt_info_s.power_consumption;
 791         for (i = 1; i < 8; ++i)
 792                 if (power_info[i].pal_power_mgmt_info_s.im
 793                     && power_info[i].pal_power_mgmt_info_s.power_consumption < min_power) {
 794                         min_power = power_info[i].pal_power_mgmt_info_s.power_consumption;
 795                         min_power_state = i;
 796                 }
 797
 798         while (1)
 799                 ia64_pal_halt(min_power_state);
 800 }
 801
 802 void
 803 machine_restart (char *restart_cmd)
 804 {
 805         (void) notify_die(DIE_MACHINE_RESTART, restart_cmd, NULL, 0, 0, 0);
 806         (*efi.reset_system)(EFI_RESET_WARM, 0, 0, NULL);
 807 }
 808
 809 void
 810 machine_halt (void)
 811 {
 812         (void) notify_die(DIE_MACHINE_HALT, "", NULL, 0, 0, 0);
 813         cpu_halt();
 814 }
 815
 816 void
 817 machine_power_off (void)
 818 {
 819         if (pm_power_off)
 820                 pm_power_off();
 821         machine_halt();
 822 }
 823