arch/x86/kernel/process_32.c

   1 /*
   2  *  Copyright (C) 1995  Linus Torvalds
   3  *
   4  *  Pentium III FXSR, SSE support
   5  *      Gareth Hughes <gareth@valinux.com>, May 2000
   6  */
   7
   8 /*
   9  * This file handles the architecture-dependent parts of process handling..
  10  */
  11
  12 #include <stdarg.h>
  13
  14 #include <linux/cpu.h>
  15 #include <linux/errno.h>
  16 #include <linux/sched.h>
  17 #include <linux/fs.h>
  18 #include <linux/kernel.h>
  19 #include <linux/mm.h>
  20 #include <linux/elfcore.h>
  21 #include <linux/smp.h>
  22 #include <linux/stddef.h>
  23 #include <linux/slab.h>
  24 #include <linux/vmalloc.h>
  25 #include <linux/user.h>
  26 #include <linux/interrupt.h>
  27 #include <linux/utsname.h>
  28 #include <linux/delay.h>
  29 #include <linux/reboot.h>
  30 #include <linux/init.h>
  31 #include <linux/mc146818rtc.h>
  32 #include <linux/module.h>
  33 #include <linux/kallsyms.h>
  34 #include <linux/ptrace.h>
  35 #include <linux/random.h>
  36 #include <linux/personality.h>
  37 #include <linux/tick.h>
  38 #include <linux/percpu.h>
  39 #include <linux/prctl.h>
  40
  41 #include <asm/uaccess.h>
  42 #include <asm/pgtable.h>
  43 #include <asm/system.h>
  44 #include <asm/io.h>
  45 #include <asm/ldt.h>
  46 #include <asm/processor.h>
  47 #include <asm/i387.h>
  48 #include <asm/desc.h>
  49 #ifdef CONFIG_MATH_EMULATION
  50 #include <asm/math_emu.h>
  51 #endif
  52
  53 #include <linux/err.h>
  54
  55 #include <asm/tlbflush.h>
  56 #include <asm/cpu.h>
  57 #include <asm/kdebug.h>
  58
  59 asmlinkage void ret_from_fork(void) __asm__("ret_from_fork");
  60
  61 DEFINE_PER_CPU(struct task_struct *, current_task) = &init_task;
  62 EXPORT_PER_CPU_SYMBOL(current_task);
  63
  64 DEFINE_PER_CPU(int, cpu_number);
  65 EXPORT_PER_CPU_SYMBOL(cpu_number);
  66
  67 /*
  68  * Return saved PC of a blocked thread.
  69  */
  70 unsigned long thread_saved_pc(struct task_struct *tsk)
  71 {
  72         return ((unsigned long *)tsk->thread.sp)[3];
  73 }
  74
  75 #ifdef CONFIG_HOTPLUG_CPU
  76 #include <asm/nmi.h>
  77
  78 static void cpu_exit_clear(void)
  79 {
  80         int cpu = raw_smp_processor_id();
  81
  82         idle_task_exit();
  83
  84         cpu_uninit();
  85         irq_ctx_exit(cpu);
  86
  87         cpu_clear(cpu, cpu_callout_map);
  88         cpu_clear(cpu, cpu_callin_map);
  89
  90         numa_remove_cpu(cpu);
  91 }
  92
  93 /* We don't actually take CPU down, just spin without interrupts. */
  94 static inline void play_dead(void)
  95 {
  96         /* This must be done before dead CPU ack */
  97         cpu_exit_clear();
  98         wbinvd();
  99         mb();
 100         /* Ack it */
 101         __get_cpu_var(cpu_state) = CPU_DEAD;
 102
 103         /*
 104          * With physical CPU hotplug, we should halt the cpu
 105          */
 106         local_irq_disable();
 107         while (1)
 108                 halt();
 109 }
 110 #else
 111 static inline void play_dead(void)
 112 {
 113         BUG();
 114 }
 115 #endif /* CONFIG_HOTPLUG_CPU */
 116
 117 /*
 118  * The idle thread. There's no useful work to be
 119  * done, so just try to conserve power and have a
 120  * low exit latency (ie sit in a loop waiting for
 121  * somebody to say that they'd like to reschedule)
 122  */
 123 void cpu_idle(void)
 124 {
 125         int cpu = smp_processor_id();
 126
 127         current_thread_info()->status |= TS_POLLING;
 128
 129         /* endless idle loop with no priority at all */
 130         while (1) {
 131                 tick_nohz_stop_sched_tick();
 132                 while (!need_resched()) {
 133
 134                         check_pgt_cache();
 135                         rmb();
 136
 137                         if (rcu_pending(cpu))
 138                                 rcu_check_callbacks(cpu, 0);
 139
 140                         if (cpu_is_offline(cpu))
 141                                 play_dead();
 142
 143                         local_irq_disable();
 144                         __get_cpu_var(irq_stat).idle_timestamp = jiffies;
 145                         pm_idle();
 146                 }
 147                 tick_nohz_restart_sched_tick();
 148                 preempt_enable_no_resched();
 149                 schedule();
 150                 preempt_disable();
 151         }
 152 }
 153
 154 void __show_registers(struct pt_regs *regs, int all)
 155 {
 156         unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L;
 157         unsigned long d0, d1, d2, d3, d6, d7;
 158         unsigned long sp;
 159         unsigned short ss, gs;
 160
 161         if (user_mode_vm(regs)) {
 162                 sp = regs->sp;
 163                 ss = regs->ss & 0xffff;
 164                 savesegment(gs, gs);
 165         } else {
 166                 sp = (unsigned long) (&regs->sp);
 167                 savesegment(ss, ss);
 168                 savesegment(gs, gs);
 169         }
 170
 171         printk("\n");
 172         printk("Pid: %d, comm: %s %s (%s %.*s)\n",
 173                         task_pid_nr(current), current->comm,
 174                         print_tainted(), init_utsname()->release,
 175                         (int)strcspn(init_utsname()->version, " "),
 176                         init_utsname()->version);
 177
 178         printk("EIP: %04x:[<%08lx>] EFLAGS: %08lx CPU: %d\n",
 179                         (u16)regs->cs, regs->ip, regs->flags,
 180                         smp_processor_id());
 181         print_symbol("EIP is at %s\n", regs->ip);
 182
 183         printk("EAX: %08lx EBX: %08lx ECX: %08lx EDX: %08lx\n",
 184                 regs->ax, regs->bx, regs->cx, regs->dx);
 185         printk("ESI: %08lx EDI: %08lx EBP: %08lx ESP: %08lx\n",
 186                 regs->si, regs->di, regs->bp, sp);
 187         printk(" DS: %04x ES: %04x FS: %04x GS: %04x SS: %04x\n",
 188                (u16)regs->ds, (u16)regs->es, (u16)regs->fs, gs, ss);
 189
 190         if (!all)
 191                 return;
 192
 193         cr0 = read_cr0();
 194         cr2 = read_cr2();
 195         cr3 = read_cr3();
 196         cr4 = read_cr4_safe();
 197         printk("CR0: %08lx CR2: %08lx CR3: %08lx CR4: %08lx\n",
 198                         cr0, cr2, cr3, cr4);
 199
 200         get_debugreg(d0, 0);
 201         get_debugreg(d1, 1);
 202         get_debugreg(d2, 2);
 203         get_debugreg(d3, 3);
 204         printk("DR0: %08lx DR1: %08lx DR2: %08lx DR3: %08lx\n",
 205                         d0, d1, d2, d3);
 206
 207         get_debugreg(d6, 6);
 208         get_debugreg(d7, 7);
 209         printk("DR6: %08lx DR7: %08lx\n",
 210                         d6, d7);
 211 }
 212
 213 void show_regs(struct pt_regs *regs)
 214 {
 215         __show_registers(regs, 1);
 216         show_trace(NULL, regs, &regs->sp, regs->bp);
 217 }
 218
 219 /*
 220  * This gets run with %bx containing the
 221  * function to call, and %dx containing
 222  * the "args".
 223  */
 224 extern void kernel_thread_helper(void);
 225
 226 /*
 227  * Create a kernel thread
 228  */
 229 int kernel_thread(int (*fn)(void *), void * arg, unsigned long flags)
 230 {
 231         struct pt_regs regs;
 232
 233         memset(&regs, 0, sizeof(regs));
 234
 235         regs.bx = (unsigned long) fn;
 236         regs.dx = (unsigned long) arg;
 237
 238         regs.ds = __USER_DS;
 239         regs.es = __USER_DS;
 240         regs.fs = __KERNEL_PERCPU;
 241         regs.orig_ax = -1;
 242         regs.ip = (unsigned long) kernel_thread_helper;
 243         regs.cs = __KERNEL_CS | get_kernel_rpl();
 244         regs.flags = X86_EFLAGS_IF | X86_EFLAGS_SF | X86_EFLAGS_PF | 0x2;
 245
 246         /* Ok, create the new process.. */
 247         return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, &regs, 0, NULL, NULL);
 248 }
 249 EXPORT_SYMBOL(kernel_thread);
 250
 251 /*
 252  * Free current thread data structures etc..
 253  */
 254 void exit_thread(void)
 255 {
 256         /* The process may have allocated an io port bitmap... nuke it. */
 257         if (unlikely(test_thread_flag(TIF_IO_BITMAP))) {
 258                 struct task_struct *tsk = current;
 259                 struct thread_struct *t = &tsk->thread;
 260                 int cpu = get_cpu();
 261                 struct tss_struct *tss = &per_cpu(init_tss, cpu);
 262
 263                 kfree(t->io_bitmap_ptr);
 264                 t->io_bitmap_ptr = NULL;
 265                 clear_thread_flag(TIF_IO_BITMAP);
 266                 /*
 267                  * Careful, clear this in the TSS too:
 268                  */
 269                 memset(tss->io_bitmap, 0xff, tss->io_bitmap_max);
 270                 t->io_bitmap_max = 0;
 271                 tss->io_bitmap_owner = NULL;
 272                 tss->io_bitmap_max = 0;
 273                 tss->x86_tss.io_bitmap_base = INVALID_IO_BITMAP_OFFSET;
 274                 put_cpu();
 275         }
 276 }
 277
 278 void flush_thread(void)
 279 {
 280         struct task_struct *tsk = current;
 281
 282         tsk->thread.debugreg0 = 0;
 283         tsk->thread.debugreg1 = 0;
 284         tsk->thread.debugreg2 = 0;
 285         tsk->thread.debugreg3 = 0;
 286         tsk->thread.debugreg6 = 0;
 287         tsk->thread.debugreg7 = 0;
 288         memset(tsk->thread.tls_array, 0, sizeof(tsk->thread.tls_array));
 289         clear_tsk_thread_flag(tsk, TIF_DEBUG);
 290         /*
 291          * Forget coprocessor state..
 292          */
 293         tsk->fpu_counter = 0;
 294         clear_fpu(tsk);
 295         clear_used_math();
 296 }
 297
 298 void release_thread(struct task_struct *dead_task)
 299 {
 300         BUG_ON(dead_task->mm);
 301         release_vm86_irqs(dead_task);
 302 }
 303
 304 /*
 305  * This gets called before we allocate a new thread and copy
 306  * the current task into it.
 307  */
 308 void prepare_to_copy(struct task_struct *tsk)
 309 {
 310         unlazy_fpu(tsk);
 311 }
 312
 313 int copy_thread(int nr, unsigned long clone_flags, unsigned long sp,
 314         unsigned long unused,
 315         struct task_struct * p, struct pt_regs * regs)
 316 {
 317         struct pt_regs * childregs;
 318         struct task_struct *tsk;
 319         int err;
 320
 321         childregs = task_pt_regs(p);
 322         *childregs = *regs;
 323         childregs->ax = 0;
 324         childregs->sp = sp;
 325
 326         p->thread.sp = (unsigned long) childregs;
 327         p->thread.sp0 = (unsigned long) (childregs+1);
 328
 329         p->thread.ip = (unsigned long) ret_from_fork;
 330
 331         savesegment(gs, p->thread.gs);
 332
 333         tsk = current;
 334         if (unlikely(test_tsk_thread_flag(tsk, TIF_IO_BITMAP))) {
 335                 p->thread.io_bitmap_ptr = kmemdup(tsk->thread.io_bitmap_ptr,
 336                                                 IO_BITMAP_BYTES, GFP_KERNEL);
 337                 if (!p->thread.io_bitmap_ptr) {
 338                         p->thread.io_bitmap_max = 0;
 339                         return -ENOMEM;
 340                 }
 341                 set_tsk_thread_flag(p, TIF_IO_BITMAP);
 342         }
 343
 344         err = 0;
 345
 346         /*
 347          * Set a new TLS for the child thread?
 348          */
 349         if (clone_flags & CLONE_SETTLS)
 350                 err = do_set_thread_area(p, -1,
 351                         (struct user_desc __user *)childregs->si, 0);
 352
 353         if (err && p->thread.io_bitmap_ptr) {
 354                 kfree(p->thread.io_bitmap_ptr);
 355                 p->thread.io_bitmap_max = 0;
 356         }
 357         return err;
 358 }
 359
 360 void
 361 start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp)
 362 {
 363         __asm__("movl %0, %%gs" :: "r"(0));
 364         regs->fs                = 0;
 365         set_fs(USER_DS);
 366         regs->ds                = __USER_DS;
 367         regs->es                = __USER_DS;
 368         regs->ss                = __USER_DS;
 369         regs->cs                = __USER_CS;
 370         regs->ip                = new_ip;
 371         regs->sp                = new_sp;
 372         /*
 373          * Free the old FP and other extended state
 374          */
 375         free_thread_xstate(current);
 376 }
 377 EXPORT_SYMBOL_GPL(start_thread);
 378
 379 static void hard_disable_TSC(void)
 380 {
 381         write_cr4(read_cr4() | X86_CR4_TSD);
 382 }
 383
 384 void disable_TSC(void)
 385 {
 386         preempt_disable();
 387         if (!test_and_set_thread_flag(TIF_NOTSC))
 388                 /*
 389                  * Must flip the CPU state synchronously with
 390                  * TIF_NOTSC in the current running context.
 391                  */
 392                 hard_disable_TSC();
 393         preempt_enable();
 394 }
 395
 396 static void hard_enable_TSC(void)
 397 {
 398         write_cr4(read_cr4() & ~X86_CR4_TSD);
 399 }
 400
 401 static void enable_TSC(void)
 402 {
 403         preempt_disable();
 404         if (test_and_clear_thread_flag(TIF_NOTSC))
 405                 /*
 406                  * Must flip the CPU state synchronously with
 407                  * TIF_NOTSC in the current running context.
 408                  */
 409                 hard_enable_TSC();
 410         preempt_enable();
 411 }
 412
 413 int get_tsc_mode(unsigned long adr)
 414 {
 415         unsigned int val;
 416
 417         if (test_thread_flag(TIF_NOTSC))
 418                 val = PR_TSC_SIGSEGV;
 419         else
 420                 val = PR_TSC_ENABLE;
 421
 422         return put_user(val, (unsigned int __user *)adr);
 423 }
 424
 425 int set_tsc_mode(unsigned int val)
 426 {
 427         if (val == PR_TSC_SIGSEGV)
 428                 disable_TSC();
 429         else if (val == PR_TSC_ENABLE)
 430                 enable_TSC();
 431         else
 432                 return -EINVAL;
 433
 434         return 0;
 435 }
 436
 437 static noinline void
 438 __switch_to_xtra(struct task_struct *prev_p, struct task_struct *next_p,
 439                  struct tss_struct *tss)
 440 {
 441         struct thread_struct *prev, *next;
 442         unsigned long debugctl;
 443
 444         prev = &prev_p->thread;
 445         next = &next_p->thread;
 446
 447         debugctl = prev->debugctlmsr;
 448         if (next->ds_area_msr != prev->ds_area_msr) {
 449                 /* we clear debugctl to make sure DS
 450                  * is not in use when we change it */
 451                 debugctl = 0;
 452                 update_debugctlmsr(0);
 453                 wrmsr(MSR_IA32_DS_AREA, next->ds_area_msr, 0);
 454         }
 455
 456         if (next->debugctlmsr != debugctl)
 457                 update_debugctlmsr(next->debugctlmsr);
 458
 459         if (test_tsk_thread_flag(next_p, TIF_DEBUG)) {
 460                 set_debugreg(next->debugreg0, 0);
 461                 set_debugreg(next->debugreg1, 1);
 462                 set_debugreg(next->debugreg2, 2);
 463                 set_debugreg(next->debugreg3, 3);
 464                 /* no 4 and 5 */
 465                 set_debugreg(next->debugreg6, 6);
 466                 set_debugreg(next->debugreg7, 7);
 467         }
 468
 469         if (test_tsk_thread_flag(prev_p, TIF_NOTSC) ^
 470             test_tsk_thread_flag(next_p, TIF_NOTSC)) {
 471                 /* prev and next are different */
 472                 if (test_tsk_thread_flag(next_p, TIF_NOTSC))
 473                         hard_disable_TSC();
 474                 else
 475                         hard_enable_TSC();
 476         }
 477
 478 #ifdef X86_BTS
 479         if (test_tsk_thread_flag(prev_p, TIF_BTS_TRACE_TS))
 480                 ptrace_bts_take_timestamp(prev_p, BTS_TASK_DEPARTS);
 481
 482         if (test_tsk_thread_flag(next_p, TIF_BTS_TRACE_TS))
 483                 ptrace_bts_take_timestamp(next_p, BTS_TASK_ARRIVES);
 484 #endif
 485
 486
 487         if (!test_tsk_thread_flag(next_p, TIF_IO_BITMAP)) {
 488                 /*
 489                  * Disable the bitmap via an invalid offset. We still cache
 490                  * the previous bitmap owner and the IO bitmap contents:
 491                  */
 492                 tss->x86_tss.io_bitmap_base = INVALID_IO_BITMAP_OFFSET;
 493                 return;
 494         }
 495
 496         if (likely(next == tss->io_bitmap_owner)) {
 497                 /*
 498                  * Previous owner of the bitmap (hence the bitmap content)
 499                  * matches the next task, we dont have to do anything but
 500                  * to set a valid offset in the TSS:
 501                  */
 502                 tss->x86_tss.io_bitmap_base = IO_BITMAP_OFFSET;
 503                 return;
 504         }
 505         /*
 506          * Lazy TSS's I/O bitmap copy. We set an invalid offset here
 507          * and we let the task to get a GPF in case an I/O instruction
 508          * is performed.  The handler of the GPF will verify that the
 509          * faulting task has a valid I/O bitmap and, it true, does the
 510          * real copy and restart the instruction.  This will save us
 511          * redundant copies when the currently switched task does not
 512          * perform any I/O during its timeslice.
 513          */
 514         tss->x86_tss.io_bitmap_base = INVALID_IO_BITMAP_OFFSET_LAZY;
 515 }
 516
 517 /*
 518  *      switch_to(x,yn) should switch tasks from x to y.
 519  *
 520  * We fsave/fwait so that an exception goes off at the right time
 521  * (as a call from the fsave or fwait in effect) rather than to
 522  * the wrong process. Lazy FP saving no longer makes any sense
 523  * with modern CPU's, and this simplifies a lot of things (SMP
 524  * and UP become the same).
 525  *
 526  * NOTE! We used to use the x86 hardware context switching. The
 527  * reason for not using it any more becomes apparent when you
 528  * try to recover gracefully from saved state that is no longer
 529  * valid (stale segment register values in particular). With the
 530  * hardware task-switch, there is no way to fix up bad state in
 531  * a reasonable manner.
 532  *
 533  * The fact that Intel documents the hardware task-switching to
 534  * be slow is a fairly red herring - this code is not noticeably
 535  * faster. However, there _is_ some room for improvement here,
 536  * so the performance issues may eventually be a valid point.
 537  * More important, however, is the fact that this allows us much
 538  * more flexibility.
 539  *
 540  * The return value (in %ax) will be the "prev" task after
 541  * the task-switch, and shows up in ret_from_fork in entry.S,
 542  * for example.
 543  */
 544 struct task_struct * __switch_to(struct task_struct *prev_p, struct task_struct *next_p)
 545 {
 546         struct thread_struct *prev = &prev_p->thread,
 547                                  *next = &next_p->thread;
 548         int cpu = smp_processor_id();
 549         struct tss_struct *tss = &per_cpu(init_tss, cpu);
 550
 551         /* never put a printk in __switch_to... printk() calls wake_up*() indirectly */
 552
 553         __unlazy_fpu(prev_p);
 554
 555
 556         /* we're going to use this soon, after a few expensive things */
 557         if (next_p->fpu_counter > 5)
 558                 prefetch(next->xstate);
 559
 560         /*
 561          * Reload esp0.
 562          */
 563         load_sp0(tss, next);
 564
 565         /*
 566          * Save away %gs. No need to save %fs, as it was saved on the
 567          * stack on entry.  No need to save %es and %ds, as those are
 568          * always kernel segments while inside the kernel.  Doing this
 569          * before setting the new TLS descriptors avoids the situation
 570          * where we temporarily have non-reloadable segments in %fs
 571          * and %gs.  This could be an issue if the NMI handler ever
 572          * used %fs or %gs (it does not today), or if the kernel is
 573          * running inside of a hypervisor layer.
 574          */
 575         savesegment(gs, prev->gs);
 576
 577         /*
 578          * Load the per-thread Thread-Local Storage descriptor.
 579          */
 580         load_TLS(next, cpu);
 581
 582         /*
 583          * Restore IOPL if needed.  In normal use, the flags restore
 584          * in the switch assembly will handle this.  But if the kernel
 585          * is running virtualized at a non-zero CPL, the popf will
 586          * not restore flags, so it must be done in a separate step.
 587          */
 588         if (get_kernel_rpl() && unlikely(prev->iopl != next->iopl))
 589                 set_iopl_mask(next->iopl);
 590
 591         /*
 592          * Now maybe handle debug registers and/or IO bitmaps
 593          */
 594         if (unlikely(task_thread_info(prev_p)->flags & _TIF_WORK_CTXSW_PREV ||
 595                      task_thread_info(next_p)->flags & _TIF_WORK_CTXSW_NEXT))
 596                 __switch_to_xtra(prev_p, next_p, tss);
 597
 598         /*
 599          * Leave lazy mode, flushing any hypercalls made here.
 600          * This must be done before restoring TLS segments so
 601          * the GDT and LDT are properly updated, and must be
 602          * done before math_state_restore, so the TS bit is up
 603          * to date.
 604          */
 605         arch_leave_lazy_cpu_mode();
 606
 607         /* If the task has used fpu the last 5 timeslices, just do a full
 608          * restore of the math state immediately to avoid the trap; the
 609          * chances of needing FPU soon are obviously high now
 610          *
 611          * tsk_used_math() checks prevent calling math_state_restore(),
 612          * which can sleep in the case of !tsk_used_math()
 613          */
 614         if (tsk_used_math(next_p) && next_p->fpu_counter > 5)
 615                 math_state_restore();
 616
 617         /*
 618          * Restore %gs if needed (which is common)
 619          */
 620         if (prev->gs | next->gs)
 621                 loadsegment(gs, next->gs);
 622
 623         x86_write_percpu(current_task, next_p);
 624
 625         return prev_p;
 626 }
 627
 628 asmlinkage int sys_fork(struct pt_regs regs)
 629 {
 630         return do_fork(SIGCHLD, regs.sp, &regs, 0, NULL, NULL);
 631 }
 632
 633 asmlinkage int sys_clone(struct pt_regs regs)
 634 {
 635         unsigned long clone_flags;
 636         unsigned long newsp;
 637         int __user *parent_tidptr, *child_tidptr;
 638
 639         clone_flags = regs.bx;
 640         newsp = regs.cx;
 641         parent_tidptr = (int __user *)regs.dx;
 642         child_tidptr = (int __user *)regs.di;
 643         if (!newsp)
 644                 newsp = regs.sp;
 645         return do_fork(clone_flags, newsp, &regs, 0, parent_tidptr, child_tidptr);
 646 }
 647
 648 /*
 649  * This is trivial, and on the face of it looks like it
 650  * could equally well be done in user mode.
 651  *
 652  * Not so, for quite unobvious reasons - register pressure.
 653  * In user mode vfork() cannot have a stack frame, and if
 654  * done by calling the "clone()" system call directly, you
 655  * do not have enough call-clobbered registers to hold all
 656  * the information you need.
 657  */
 658 asmlinkage int sys_vfork(struct pt_regs regs)
 659 {
 660         return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, regs.sp, &regs, 0, NULL, NULL);
 661 }
 662
 663 /*
 664  * sys_execve() executes a new program.
 665  */
 666 asmlinkage int sys_execve(struct pt_regs regs)
 667 {
 668         int error;
 669         char * filename;
 670
 671         filename = getname((char __user *) regs.bx);
 672         error = PTR_ERR(filename);
 673         if (IS_ERR(filename))
 674                 goto out;
 675         error = do_execve(filename,
 676                         (char __user * __user *) regs.cx,
 677                         (char __user * __user *) regs.dx,
 678                         &regs);
 679         if (error == 0) {
 680                 /* Make sure we don't return using sysenter.. */
 681                 set_thread_flag(TIF_IRET);
 682         }
 683         putname(filename);
 684 out:
 685         return error;
 686 }
 687
 688 #define top_esp                (THREAD_SIZE - sizeof(unsigned long))
 689 #define top_ebp                (THREAD_SIZE - 2*sizeof(unsigned long))
 690
 691 unsigned long get_wchan(struct task_struct *p)
 692 {
 693         unsigned long bp, sp, ip;
 694         unsigned long stack_page;
 695         int count = 0;
 696         if (!p || p == current || p->state == TASK_RUNNING)
 697                 return 0;
 698         stack_page = (unsigned long)task_stack_page(p);
 699         sp = p->thread.sp;
 700         if (!stack_page || sp < stack_page || sp > top_esp+stack_page)
 701                 return 0;
 702         /* include/asm-i386/system.h:switch_to() pushes bp last. */
 703         bp = *(unsigned long *) sp;
 704         do {
 705                 if (bp < stack_page || bp > top_ebp+stack_page)
 706                         return 0;
 707                 ip = *(unsigned long *) (bp+4);
 708                 if (!in_sched_functions(ip))
 709                         return ip;
 710                 bp = *(unsigned long *) bp;
 711         } while (count++ < 16);
 712         return 0;
 713 }
 714
 715 unsigned long arch_align_stack(unsigned long sp)
 716 {
 717         if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space)
 718                 sp -= get_random_int() % 8192;
 719         return sp & ~0xf;
 720 }
 721
 722 unsigned long arch_randomize_brk(struct mm_struct *mm)
 723 {
 724         unsigned long range_end = mm->brk + 0x02000000;
 725         return randomize_range(mm->brk, range_end, 0) ? : mm->brk;
 726 }